Patent Translate Powered by EPO and Google Notice This translation is machine-generated. It cannot be guaranteed that it is intelligible, accurate, complete, reliable or fit for specific purposes. Critical decisions, such as commercially relevant or financial decisions, should not be based on machine-translation output. DESCRIPTION JPH0795684 [0001] BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an acoustic characteristic correction apparatus for correcting response characteristics (such as frequency response) of a reproduction system including a sound field such as a listening room to a desired characteristic. It is [0002] 2. Description of the Related Art Conventionally, graphic equalizers have been generally used as a device for correcting the response characteristics of the entire reproduction system including a room, a speaker and the like. This is to divide the voice frequency band into several bands and adjust the gain for each divided band. However, with this, it was not possible to know how to adjust the playback sound to have the desired response characteristics. [0003] In order to solve the drawbacks of the conventional graphic equalizer and to automatically set the response characteristic of the whole reproduction system to the desired characteristic, for example, there is the one described in Japanese Patent Publication No. 61-59004. This allows the user to set desired characteristics and reproduces measurement signals such as white noise and impulses in the sound field to be reproduced by the speaker of the reproduction system, 08-05-2019 1 collecting this with a microphone, and responding to the response characteristics Measure the correction characteristic so that it matches the desired characteristic, set the filter characteristic of the equalizer that matches this correction characteristic, and reproduce the music signal through this equalizer to adjust it to the desired characteristic It is intended to make music playback enjoyable. [0004] In the above-described conventional apparatus, a configuration for measuring the response characteristic and a configuration for correcting the response characteristic based on the measurement result are separately required, and a hardware configuration is required. It was getting larger. [0005] The present invention has been made in view of the above-described points, and an object of the present invention is to provide an acoustic characteristic correction device in which the device can be miniaturized by simplifying the configuration of the device. [0006] The invention according to claim 1 is an inverse filter for measuring response characteristics when using a TSP (Time Stretched Pulse) signal as a measurement signal. A common convolution calculator is used to apply time compression and correction characteristics by characteristics. [0007] According to the second aspect of the invention, when the convolution operator is shared in this way, the convolution operator has the number of stages necessary for giving the correction characteristic and does not have the number of stages necessary for time compression by the inverse filter characteristic. , And the time compression is performed in time divisions. [0008] According to the first aspect of the present invention, when the response characteristic is measured using the TSP signal as the measurement signal, a convolution common to time compression and provision of the correction characteristic by the inverse filter characteristic at the time of measurement Since the arithmetic unit is used, the hardware configuration is simplified, and the apparatus can be miniaturized. 08-05-2019 2 [0009] When this is realized, generally, the number of stages of the convolution unit necessary for time compression by the inverse filter is often much larger than the number of stages necessary for giving the correction characteristic, and the number of stages necessary for time compression by the inverse filter is prepared Then, it is considered to be useless for the provision of the correction characteristic. Therefore, according to the second aspect of the present invention, the number of stages of the convolution operation unit is the number of stages required for giving the correction characteristic, and the time reduction is performed by dividing the time by the inverse filter characteristic. Is used to share time compression and correction characteristics. [0010] An embodiment of the present invention will be described below. FIG. 2 shows an outline of the hardware configuration of the entire apparatus. The acoustic characteristic correction device 10 is composed of a main body 12 and a remote controller 14, and the both are connected by a detachable signal cable 16. [0011] When measuring the response characteristic, the main unit 12 measures the generation of the measurement signal, calculates the frequency characteristic based on the microphone pickup signal, calculates the correction characteristic, calculates the FIR filter coefficient corresponding to the correction characteristic, etc. When the latter is used as an equalizer, the response characteristic is corrected by applying a set FIR filter characteristic to the acoustic signal to be reproduced. The remote control unit 14 displays instructions of various operations at the time of 08-05-2019 3 measurement and desired characteristics setting and displays various response characteristics (measurement characteristics, desired characteristics, correction characteristics, etc.) on the main unit 12. [0012] In the main body 12, the measurement microphone is connected to the microphone input terminal 18 at the time of measurement of the response characteristic, and a microphone sound collection signal is input. Further, a source device such as a CD player is connected to the source input terminal 20, and a source signal reproduced from the source device at the time of use as an equalizer is input. The input unit 22 performs A / D conversion of microphone input and source input. The output unit 24 D / A converts the equalized source signal and the measurement signal (test tone signal) and outputs the converted signal from the output terminal 26. The patch bay unit 28 changes the connections of input and output and various other signals between measurement and equalizer. The waveform memory output unit 30 reads out and outputs the measurement signal waveform (band signal waveform, TSP signal waveform) and TSP inverse filter waveform stored in the ROM. [0013] The input waveform memory unit 32 stores the A / D converted microphone input in the RAM. The convolution operation unit 34 is composed of a real-time convolution circuit (for example, a circuit in which thousands of stages (for example, 4000 to 8000 stages) convolvers are connected by cascade connection of LSIYM 7309 manufactured by Yamaha Corporation). By transmitting the filter coefficients of the equalizer here, the equalizer with the FIR filter is configured. Further, at the time of measurement in the case of using the TSP signal as a measurement signal, the TSP inverse filter coefficient is transferred to the convolution unit 34 to configure the TSP inverse filter. Data processing calculation Other control unit 36 is constituted by a CPU, processing of measurement data (measurement characteristic, desired characteristic, calculation of correction characteristic, calculation of equalizer filter coefficient corresponding to correction characteristic (Fourier inverse transform), etc.), patch bay unit 28 And other necessary control of the main unit 12 and exchange of signals with the CPU 42 of the remote control unit 14. 08-05-2019 4 [0014] In the remote control unit 14, the operation unit 38 issues all instructions necessary for the main unit 12 at the time of measurement, desired characteristics setting, and correction characteristics setting. The display unit 40 displays various response characteristics and displays for operation. The CPU 42 exchanges data with the CPU 36 of the main unit 12. [0015] An example of the panel configuration of the remote control unit 14 is shown in FIG. The display unit 40 is configured of an LCD display or the like, and various response characteristics are graphically displayed. That is, in the upper graph display section, the measurement characteristic is a bar graph 44 on the common graph axis (horizontal axis is frequency, vertical axis is level), and an example of the desired characteristic (flat characteristic is shown). ) Are displayed superimposed on the line graph 46. In addition, cursors 62 and 64 for indicating the frequency range are displayed by vertical lines. In the lower part, a correction graph calculated as a difference between the desired characteristic and the measurement characteristic is displayed as a line graph 48. In addition, the correction frequency range set by the operation of the operator is displayed as a horizontal bar graph 50 between the upper and lower stages. In this case, the correction characteristic display 48 is not displayed outside the correction frequency range (or displayed flat on the 0 dB line). In the upper and lower portions of the graph display unit, display portions 52 and 54 for displaying current setting items, setting contents and the like are provided to assist the operator's operation. [0016] In the operation unit 38, a cursor key 56, a shuttle key (rotary encoder) 58, various key switches 60, and the like are provided. The cursor key 56 includes an up key 56a, a down key 56b, a left cursor selection key 56c, and a right cursor selection key 56d. The left and right cursor selection keys 56c and 56d are used to select one of the left and right cursors 62 and 64 of the display unit 40, for example, when correcting a desired characteristic or setting a correction frequency range. When the left cursor selection key 56c is pressed and the shuttle 58 is turned, the left cursor 62 is moved in the turned direction to set the lower limit value of the frequency range. When the right cursor selection key 56d is pressed and the shuttle 58 is turned, the right cursor 64 moves in the turning direction, and the upper limit value of the frequency range is set. For 08-05-2019 5 example, a ▽ mark 65 is displayed at the position of the selected one of the cursors 62 and 64, which makes it possible to know which one is selected. The up and down keys 56a and 56b are used, for example, to correct the desired characteristic. When the up key 56a is pressed for the designated frequency range, the level of the desired characteristic is gradually raised in a curve, and the down key 56b is used. When pressed, the level of the desired characteristic is gradually lowered in a curve. The key switch 60 is used for selection of setting items, selection of measurement data, execution instructions, and various other instructions. [0017] An outline of a procedure from measurement of frequency characteristics to use as an equalizer using the acoustic characteristic correction device 10 of FIG. 2 is shown in FIG. Each process is sequentially advanced by mode progress operation by the operator (for example, every time one key switch is pressed, it progresses to the next process). The outline of each process will be described. [0018] Test As shown in FIG. 5A, a microphone 72 is disposed at a listening position 71 in a room 70 where music is reproduced, a measuring signal is outputted from the unit 10, and a speaker used for reproduction through a power amplifier 74. The sound is reproduced from 76 and 78, collected by the microphone 72, and the sound-wave form is taken into the memory in the machine 10. This measurement is performed at each position by moving the microphone 72 to a plurality of points (for example, five points) centered on the receiving position 71 as shown on the right of FIG. 5A, as necessary. [0019] The response characteristic is calculated based on the collected sound signal captured in the calculation memory of the measurement characteristic. The obtained response characteristic (measurement characteristic) is displayed on the display unit 40 of the remote control unit 14 as a bar graph as shown in FIG. 6A, for example. [0020] 08-05-2019 6 Setting of Desired Characteristics The remote controller 14 operates the operation unit 38 while looking at the display unit 40 to set desired characteristics. The desired characteristic selected or set is displayed on the same graph axis as the display 44 of the measurement characteristic on the display unit 40 as shown in FIG. When desired characteristics are set such that the measured characteristics 44 are smoothed and flat as shown in FIG. 6B, for example, both characteristic displays are displayed superimposed on the same graph axis. You can see at a glance what kind of desired characteristics you want to be flat and it is easy to set. [0021] Calculation desired characteristics of the correction characteristics are set, the correction characteristics are automatically calculated as the difference between the desired characteristics and the measurement characteristics, and displayed on the display unit 40 as a line graph 50 as shown in FIG. Even when the desired characteristic is corrected, the correction characteristic is calculated at any time and displayed. [0022] If the peak of the correction characteristic of the correction characteristic is large, the sense of incongruity may occur, so the upper and lower limit values of the level of the correction characteristic are regulated as necessary. Further, when the correction range is limited due to the limitation of the reproduction frequency characteristics of the speaker to be used, the correction frequency range is restricted as needed (that is, the correction amount outside the correction frequency range is made 0 dB). [0023] When the calculation correction characteristic of the equalizer filter coefficient is determined, it is subjected to inverse Fourier transform to obtain a corresponding impulse response. In this case, it is arbitrarily selected and used from linear phase processing inverse Fourier transform, minimum phase processing inverse Fourier transform, or Fourier inverse transform of another algorithm according to the use condition etc. As a result, an impulse response as shown in FIG. 6 (d) or (e) is obtained. An equalizer (FIR) filter coefficient is given as a level value at each position 08-05-2019 7 on the time axis of this impulse response. In this way, the equalizer characteristics over the entire frequency band are set. [0024] Confirmation of correction effect Confirm the correction effect as necessary. This sets the obtained equalizer filter coefficient in the convolution unit 34 to form an equalizer, applies correction characteristics to the measurement signal with this equalizer, reproduces from the speaker, measures the response characteristics again, and displays The measured characteristic and the desired characteristic are displayed superimposed on the part 40 to confirm the correction effect. As the two characteristics match, the correction according to the desired characteristics is performed. If the expected correction state can not be obtained due to the limitations of the speaker characteristics, etc., the desired characteristics are re-corrected as necessary. [0025] When the music reproduction equalizer filter characteristics are finally determined, as shown in FIG. 5B, the source device 80 such as a CD player is connected and the main unit 12 of the main unit 10 is used as an equalizer to achieve the final purpose. Play music. [0026] FIG. 1 shows a control block configuration in the acoustic characteristic correction device 10 for realizing each step of the above procedure. FIG. 1 shows the connection state at the time of measurement. The measurement microphone 72 and the source device 80 are connected to the microphone input terminal 18 and the source input terminal 20, respectively. The measurement signal input from the microphone input terminal 18 is amplified by the microphone amplifier 82. The switch 84 is switched between measurement and calculation (the above steps) and regeneration (the above steps). A / D converter 86 converts the microphone input or analog source input into a digital signal. The switch 88 is for passing the digital source input to the bypass 90, and is switched between the digital source input reproduction and the other modes (analog input reproduction, measurement). The switch 92 is switched between measurement and reproduction. The waveform memory 32 takes in the microphone input at the time of test. The measurement signal 08-05-2019 8 generator 30 is configured of a ROM that stores the waveform of the measurement signal. In this embodiment, a band signal of band signal method (to be described later) and a TSP signal of TSP method (to be described later) are stored as measurement signals, and any one of them can be read out according to the selection operation of the operator. It has been [0027] The switch 94 is switched at the time of reproduction, at the time of response characteristic calculation, and at the time of test. The switch 96 switches between the route through the convolution unit 34 and the route 98 for bypassing the same. In the test and in the response characteristic calculation in the band signal method, the bypass path 98 is selected, and the response characteristic calculation in the TSP method is corrected At the time of confirmation of the effect and at the time of music reproduction, a route passing through the convolution unit 34 is selected. The application of the convolution operator 34 can be switched by switching the switch 102. That is, at the time of response characteristic calculation in the TSP method, the TSP inverse filter waveform read out from the TSP inverse filter waveform memory 100 is set as a filter coefficient, time-compacts the collected TSP signal as a TSP inverse filter, and impulse response Ask. In addition, at the time of confirmation of the correction effect and at the time of music reproduction (at the time of addition of the correction characteristic), an equalizer filter coefficient corresponding to the correction characteristic obtained by calculation is set as a filter coefficient to operate as an equalizer. As a result, since the convolution operation unit 34 is used both as an inverse filter in the TSP method at the time of response characteristic operation and at the time of correction effect confirmation and at the time of music reproduction, the hardware configuration is simplified. There is no problem at all in this way because the response characteristic calculation, the correction effect confirmation and the music reproduction are not performed simultaneously. [0028] When the convolution unit 34 has the number of stages necessary for giving the correction characteristic but does not have the number of stages necessary for time compression due to the inverse filter characteristic, time compression is performed by dividing the time. [0029] The output of the convolution unit 34 or the output through the bypass 98 is input to the switch 106 through the summing point 104. 08-05-2019 9 The switch 106 is switched between a test, a correction effect confirmation, a music reproduction and a response characteristic calculation. At the time of test, at the time of correction effect confirmation, at the time of music reproduction, the measurement signal or music signal passed through the switch 106 is converted into an analog signal by the D / A converter 108 and the low pass filter 110 and output from the output terminal 26 It is reproduced by the speakers 76 and 78 in the room 70 through the power amplifier 74. [0030] The signal led from the switch 106 to the line 112 at the time of response characteristic calculation is distributed by the switch 114 according to the measurement method. That is, in the case of the TSP method, after the impulse response signal is Fourier-transformed by the frequency conversion means 116 and converted into frequency information, the band division means 118 divides it into predetermined frequency bands (for example, every 1/3 octave band). Further, in the case of the band signal method, since the measurement data is obtained in the state of being originally divided into frequency bands (for example, every 1/3 octave band), it is passed through the bypass path 120 as it is. The signals of both paths pass through a summing point 122 and a band power averaging circuit 124 obtains a power average for each divided band. The band power data of the entire frequency band determined is stored in the band data memory 126. The band data memory 126 can store measurement data of a plurality of times (for example, eight times). The measurement data of each time is displayed as a bar graph according to the display selection operation of the operator (measurement characteristic display 44 of FIG. 3). [0031] The selection / weighting means 128 selects and outputs a plurality of measurement data stored in the band data memory 126 which are selected and instructed by the operator's selection operation. The measurement data is weighted according to the positions of the measurement points P1 to P5 (FIG. 5A) with respect to the listening position 71 as required. The collective averaging means 130 calculates a collective average of the plurality of selected and weighted measurement data. Interpolation means 132 treats the values of each band averaged in a group as the value at the center frequency of each band, interpolates between the center frequencies of each band, and connects all frequency bands with continuous smooth curve data Find the characteristics. The interpolation data thus obtained is stored in the RAM 134 as a final 08-05-2019 10 measurement characteristic. [0032] The ROM 136 stores average characteristics and other characteristics as desired characteristics, and the one selected by the key switch 60 is read out. The selected desired characteristic is corrected to the desired characteristic by the computing means 140 based on the operation of the cursor key 56, the shuttle key 58 and the like by the operator. The corrected desired characteristics are stored in the backup power-supply RAM 138, and can be read out and used as needed as the characteristics of the ROM 136. [0033] The calculating means 142 calculates the correction characteristic from the set desired characteristic and the measured characteristic. As the correction characteristics, corrections such as upper / lower limit regulation of the correction level and regulation of the correction frequency range are added based on the operation of the operator as necessary. The equalizer filter coefficient calculation means 144 calculates an equalizer filter coefficient corresponding to the set correction characteristic. The calculated filter coefficient is set in the convolution unit 34, and the equalizer characteristic at the time of correction effect confirmation is set at the time of music reproduction. Also, the calculated filter coefficient is stored in the backup power supply RAM 146 and can be read out and used as needed. It is also stored in the RAM card 148, and this filter coefficient can be shared by inserting the RAM card 148 into another music characteristic correction device. [0034] The display control means 150 performs control to display the calculated measurement characteristics, desired characteristics, correction characteristics and the like on the display unit 40 of the remote control unit 14. The CPU 36 (FIG. 2) of the main unit 12 executes various operations other than the switching control of each switch in FIG. 1 and the convolution operation unit 34. [0035] 08-05-2019 11 Next, control of each process of the procedure of FIG. 4 by the control block of FIG. 1 described above will be described in detail. When measuring the response characteristics indoors, the characteristics differ considerably depending on the location. This is because the reflected waves from the ceiling, floor, wall and the like in the room interfere with each other to disturb the frequency characteristics. In addition, this phenomenon is more pronounced at short highfrequency short wavelength differences. Therefore, if the correction coefficient is determined based on the data of one measurement point to obtain the filter coefficient of the equalizer, the best result is given at that point, but the area including the periphery (the range where the head of the listener moves etc. In some cases, extreme peak dips may occur and the best result may not be obtained. [0036] Therefore, in this embodiment, as shown on the right of FIG. 5A, a measurement area 73 centered on the listening position 71 in the room 70 is set, and the listening position 71 is included in the area 73. A plurality of measurement points P1 to P5 are set, the microphone 72 is moved to each of the points P1 to P5 to perform measurement, and a correction characteristic is obtained from the spatial average of them. As a result, good correction characteristics can be obtained on average at any position within the area, and the effective area of correction can be expanded. [0037] Further, in this embodiment, as the test method, either one of the band signal method and the TSP method can be selected according to the selection operation of the operator as described above. The TSP method has an advantage that measurement time can be short, and continuous measurement data can be obtained instead of discrete measurement data for each divided band. However, in this embodiment, as described above, since the convolution operator 34 for the equalizer is also used as the TSP inverse filter used in the TSP method, the length of the TSP signal for measurement is limited. The power of the entire TSP signal is limited, and when used for measurement in a noisy environment, the SN ratio of the measurement result may be degraded. [0038] 08-05-2019 12 Therefore, band signal method is used when there is no restriction in noise environment or measurement time, and when noise environment is small or measurement time is limited (for example, reproduction system (speaker system in hall etc. If there are many cases, and if it takes a lot of time to measure with band signal method, etc.), use TSP method and use both methods properly. [0039] The band signal method and the test method using the TSP method will be respectively described. (A) Band signal method The band signal method is to measure the response of each band by sequentially emitting a band signal having a plurality of frequency bands divided at different times. Here, as the bandwidth of each band, a 1/3 octave band method (that is, a division method in which each band has a 1/3 octave bandwidth) which is said to be relatively close to the auditory characteristic is used. In this case, although it is possible to obtain continuous data with high division ability if the division pitch is finely taken, it takes a huge amount of time to emit band signals of the entire band. Therefore, here, the division pitch is set by the operator's selection operation to either every 1/3 octave in FIG. 7A or every 1/6 octave in FIG. 7B, and the measurement data is measured. Interpolation is performed to obtain continuous data. If the division pitch is 1/3 octave pitch, the bandwidth does not overlap, and if it is 1/6 octave pitch, the bandwidth shifts while overlapping by 1/3 octave. If overlapping is made, the connection between the bands in the measurement data becomes good. [0040] FIG. 8 shows an example in the case of dividing into 1/3 octave pitch by 1/3 octave band. (A) is a band signal waveform center frequency, (b) is a band signal waveform (when the center frequency is 100 Hz), and (c) is an output flow of the band signal waveform. The band signal waveform of (b) is stored in the measurement signal generator 30 (ROM) of FIG. 1, and the measurement signal of each band is generated by changing the reading speed. Band signals sequentially emitted from the speakers 76 and 78 at different times are collected by the microphone 72 for each band, and the sound wave form is stored in the waveform memory 32 of FIG. 08-05-2019 13 [0041] (B) TSP method In general, a single pulse is used to measure an impulse response such as a hole, but since the signal power is small, a sufficient signal-to-noise ratio often can not be obtained even by using methods such as synchronous addition. . On the other hand, when the TSP signal is used, the signal power is large and the SN ratio can be easily obtained. In addition, since an inverse filter can be easily obtained and the response of the TSP signal can be converted into an impulse response, the convolution operation with the inverse filter may be performed. Therefore, the TSP signal has convenient characteristics for measurement. [0042] The TSP signal used in the TSP method has a waveform as shown in FIG. 9 (a). The TSP waveform is stored in the measurement signal generator 30 shown in FIG. 1, read out once from one measurement, and reproduced from the speakers 76 and 78. The reproduced TSP signal is collected by the microphone 72, and the sound wave form is stored in the waveform memory 32. [0043] The calculation of the response characteristic based on the sound wave form stored in the arithmetic waveform memory 30 of the measurement characteristic is performed as follows according to the test method. [0044] (A) Band signal method In the band signal method, the sound absorption type for each divided band stored in the waveform memory 30 of FIG. 1 is instantly switched to switches 94 and 96, bypass 98, summing point 104, switches 106 and 114, After passing through the bypass path 120 and the addition point 122, the band power average calculation unit 124 calculates the band power average for each divided band and stores it in the band data memory 126. The band data memory 126 can store measurement data for a plurality of times, and stores, for example, measurement data of five points P1 to P5 shown on the right of FIG. 5A. In the selection weighting means 128, the operator looks at individual measurement characteristics on the 08-05-2019 14 display unit 40, and discards data, for example, by excluding data extremely different from the others. Also, the remaining data is weighted as necessary. Specifically, in the case where the measurement points are, for example, 5 points P1 to P5 shown on the right in FIG. 5A, weighting is performed using the point P1 at the center position (position where the head is mainly) as 1 Each of P2 to P5 is set to 0.5, and the point P1 at the center position is set to 1, and the other points P2 to P5 are totaled to 1 and so on. [0045] The selected and weighted measurement data is subjected to collective averaging in collective averaging means 130. This gives an average measurement data of the area where the measurement was made. Since the aggregate averaged measurement data is discrete data for each divided band, it is interpolated by the interpolation means 132 to be converted into continuous smooth curve data. As the interpolation method, a spline interpolation method capable of interpolation in a short time is suitable. Interpolation treats data obtained as power average for each divided band as a value at the center frequency of each band as shown in FIG. 10, and spline interpolates between each point based on the values of several points before and after Then, for example, interpolation data of 4096 points is obtained and used as a measurement characteristic. [0046] In this way, by obtaining the power average for each divided band and using it as the value at the center frequency, spline interpolation is performed between each point to achieve useful and practical averaging of the obtained measurement characteristic results. As described above, it is possible to prevent the occurrence of a large peak and dip due to phase interference in the measurement characteristics, so that the sense of incongruity due to extreme correction is prevented when the correction characteristics are obtained using the measurement characteristics as they are. Ru. The data of the measurement characteristic thus obtained is stored in the RAM 134 of FIG. 1 and displayed on the display unit 40 as a bar graph (measurement characteristic display 44 of FIG. 3). [0047] (B) TSP method The sound wave form stored in the waveform memory 32 of FIG. 1 in the TSP 08-05-2019 15 method is the reverse TSP waveform stored in the TSP inverse filter coefficient memory 100 by the convolution operator 34 via the switches 94 and 96 immediately. A convolution operation (time compression) is performed with (FIG. 9 (b)) to obtain an impulse response (FIG. 9 (c)). The reverse TSP waveform is a waveform obtained by temporally inverting the TSP waveform (FIG. 9A). When the number of stages of the convolution operation unit 34 is insufficient as a time compression filter of the TSP signal, time compression can be divided and performed as described above. [0048] The impulse response output from the convolution unit 34 passes through the summing point 104 and the switches 106 and 114, and is Fourier-transformed by the frequency converter 116 to obtain the frequency response characteristic (FIG. 9 (d)). The determined frequency response characteristic is band-divided by the band division means 18 into the same state (1/3 octave bandwidth, 1/3 or 1/6 octave pitch) as the band signal method. The band-divided measurement data is subjected to the same processing as in the case of the band signal method. That is, the measurement data of each band divided by the band dividing means 118 passes through the addition point 122, the band power average of each divided band is calculated by the band power average calculating means 124, and stored in the band data memory 126. Ru. The band data memory 126 stores measurement data of multiple points and a plurality of times. In the selection weighting means 128, the operator looks at individual measurement characteristics on the display unit 40, and discards data, for example, by excluding data extremely different from the others. Also, the remaining data is weighted as necessary. The selected and weighted measurement data is subjected to collective averaging in collective averaging means 130. This gives an average measurement data of the area where the measurement was made. Since the aggregate averaged measurement data is discrete data for each divided band, this is splineinterpolated by the interpolation means 132 to be converted into continuous smooth curve data. The interpolated measurement data is stored in the RAM 34 as a measurement characteristic and displayed on the display unit 40 as a bar graph. [0049] As described above, in the TSP method as well, measurement data is divided once into bands, power averaging is performed for each band, and interpolation is performed to obtain continuous data. As a result, it is possible to prevent a sense of incongruity due to extreme correction when the correction characteristic is obtained using the measurement characteristic as it is and used for the characteristic correction. 08-05-2019 16 [0050] Here, a specific example of a method of dividing time compression will be described. FIG. 22 shows the hardware configuration. As for input data (sound collection data), a part necessary for the convolution operation with the reverse TSP waveform is sequentially stored in the temporary buffer (RAM) 152 via the control unit 150. A coefficient memory (TSP inverse filter waveform memory) 100 stores a TSP inverse filter waveform as a coefficient value of convolution operation. For example, as shown in FIG. 23, the convolution operation unit 34 has an input data register 154 for holding a plurality of input data, a coefficient register 156 for holding a plurality of coefficient data to be associated with the respective input data, It comprises a multiplier 158 which sequentially multiplies each coefficient data and an accumulator 160 which accumulates each multiplied value. If the number of stages of the registers 154 and 156 of the convolution unit 34 is less than the number of stages required for the convolution operation with the reverse TSP waveform, it can be simply dealt with by cascading the convolution unit 34 of FIG. . However, in this case, the number of stages of the convolution operation unit 34 is greatly increased, and the effect of using the convolution operation unit 34 for giving the correction characteristic is lost. [0051] Therefore, in the configuration of FIG. 22, the convolution with the inverse TSP waveform is divided into a plurality of times each possible by one convolution operation unit 34 and time division is performed, and the accumulation results of each time are summed up to obtain the final result. Calculating the convolution value. That is, the control unit 150 should read out data corresponding to the number of data that can be subjected to the convolution operation at one time out of the input data stored in the temporary buffer 152 and add it to each data read out from the coefficient memory 100. The coefficients are read out and subjected to a convolution operation, and the calculation result (the accumulation result up to the middle) is temporarily stored in the temporary buffer 152. Subsequently, a convolution operation is similarly performed on the data of the next divided portion, and the operation result is added to the previous accumulation result. In this manner, the final calculation result can be obtained by repeating the convolution operation for each divided portion and the addition with the accumulated value up to that point. [0052] 08-05-2019 17 Specifically, if it is necessary to perform an operation to obtain one output sample y (x), the operation is divided into m times (however, the number of possible convolution operations is 1: 1 time) ). [0053] FIG. 24 shows a control block configuration by the control unit 150 for realizing the abovedescribed divided convolution. The temporary buffer 152 has an area for storing input data and an area for storing an accumulated value. The input data is stored in the temporary buffer 152 via the control means 162. The control means 162 reads out input data of one divided portion from the temporary buffer 152, reads out from the coefficient memory 100 coefficient data to be associated with this, and performs a convolution operation in the convolution operation unit 34. Further, the accumulated result up to that time is read out from the temporary buffer 152 and added to the operation value at that time by the convolution unit 34 to obtain a new accumulated value. When a new accumulated value is obtained, the accumulated value of the temporary buffer 152 is updated to this new accumulated value. Then, by repeating the convolution operation for each divided portion, addition with the accumulated value up to that point, and updating of the accumulated value, the final accumulated value is obtained. The accumulated value is read out from the temporary buffer 152 as a final calculation result and output through the control means 162. FIG. 25 shows a flowchart of the above control, and control is performed such that this control flow is completed within one sample period of input data. [0054] Another example of control by the control unit 150 is shown in FIG. This configuration is such that accumulation of the operation value for each divided portion is performed in the control means 164 without being performed in the convolution unit 34. That is, when the operation value of one divided portion is obtained, the control means 164 reads out the accumulated value up to that point from the temporary buffer 152 and adds them, and the added value is taken as a new accumulated value. Update calculated data. The final accumulation result is read from the temporary buffer 152 and output through the control means 162. [0055] 08-05-2019 18 An example of the operation procedure in the calculation of the test and the measurement characteristic described above is shown in FIG. First, the microphone position is set (S1), and either the band signal method or the TSP method is selected as a test method (S2). Further, either 1/3 octave band pitch or 1/6 octave band pitch is selected as the band division pitch (S3). Thereafter, when the test start button is pressed (S4), a test sound is reproduced from the speakers 76 and 78, collected by the microphone 72, and stored in the waveform memory 32 (S5). The measurement results are immediately displayed on the display unit 40 as a bar graph (S6), and the operator can see and confirm this. If the measurement result seems abnormal (for example, large noise etc.), re-testing is performed at that point (S7, S8). If the measurement result is good, the microphone position is moved to another point and the test is repeated (S9). [0056] When the test is completed for all points (S10), collected data are sequentially displayed on the display unit 40, and data selection is performed as needed (S11). The selected data is weighted for each measurement point by automatic or manual setting as required (S12). Then, the aggregate average value and further the interpolation value are automatically calculated for the data of each point that has been weighted, and stored as one final measurement characteristic data in the RAM 134 (S13), and the measurement is ended. [0057] Setting of Desired Characteristics One example of a setting flow of desired characteristics is shown in FIG. When the desired characteristic setting mode is selected and operated by the remote control unit 14 (FIG. 3), a graph scale is displayed on the display unit 40 (S22), and the measurement characteristic stored in the RAM 134 is displayed by the bar graph 44 (S23). Next, when a desired characteristic selection operation is performed (S24), the corresponding desired characteristic is read from the ROM 136 or the RAM 138 and displayed on the display unit 40 by the line graph 46 (S25). [0058] By the way, the transmission characteristic of the speaker in the listening room or the hall 08-05-2019 19 changes depending on the directivity of the speaker and the reverberation characteristic of the room, and the desired characteristic of the sense of hearing does not always coincide with the flattening of the measurement characteristic. Therefore, it is convenient if the desired characteristics in the room can be easily set. For example, in a large speaker system, it is easily possible by preparing in advance a desired characteristic desired as a characteristic for PA (Public Address) in a hall or a desired characteristic when listening with a small speaker in a home listening room. Correction to that characteristic is possible. [0059] Therefore, the ROM 136 has, as a general pattern of desired characteristics, for example, a flat characteristic C1 over the entire band as shown in FIG. It is convenient to prepare in advance the medium sound emphasis characteristic C4, the low / high sound emphasis characteristic C5 and the like. In this case, by displaying the characteristic pattern name on the display unit 40, the operator refers to it and moves the cursor to the desired characteristic pattern to perform selection operation, and reads out the corresponding characteristic data from the ROM 136 to obtain the desired characteristic. It can be used as In addition, the characteristic data classified in various speakers (in-hall PA, outdoor PA, small speaker for studio monitor, etc.) and various rooms (Japanese-room listening room, Western-room listening room, etc.) are stored in the ROM 136. By displaying the speaker type name and the room type name on the screen, the operator refers to this and moves the cursor according to the speaker type and room to be used to select and operate the speaker type and the room type. Characteristic data can be read from the ROM 136 and used as a desired characteristic. When the desired characteristic is set, the calculation means 142 automatically calculates [measurement characteristic] − [desired characteristic] to obtain the correction characteristic, which is displayed on the display unit 40 by the line graph 48 (see FIG. S27). The characteristic data read out from the ROM 136 can be used as it is as a desired characteristic, but can also be partially corrected and used. [0060] As a characteristic adjustment method in the case of the conventional graphic equalizer and parametric equalizer, as shown in FIG. 14, it is common to adjust by changing the values of the center frequency F, the gain G and the sharpness Q. In this case, as the order of adjustment, the center frequency F is determined first, then the value of Q is set, and finally the gain G is increased or decreased. The adjustment operation was not easy because it had to be adjusted to the characteristics. In addition, when Q is changed, the influence extends to the entire frequency range, so it was difficult to grasp how the characteristics actually change when Q was changed, 08-05-2019 20 and it was difficult to adjust. [0061] Therefore, here, instead of determining the center frequency, the frequency range is set from where to where, and by keeping the smooth connection of the characteristics at both ends up and down the characteristics within the specified range, it is smooth and human sense It is possible to easily set the characteristic curve of the desired characteristic close to. The steps following step S28 in FIG. 12 showing this setting procedure will be described. [0062] Initially, one of the frequency range lower limit value or the upper limit value designated by the cursors 62 and 64 on the display unit 40 is selected and can be corrected (the selected one is selected). ▽ mark 65 is displayed. When the shuttle key 58 is operated in this state (S28), the selected one of the frequency range upper limit value and the lower limit value changes in the direction in which the shuttle key 58 is turned (S29, S30, S31), The cursor 62 or 64 having the ▽ mark 65 on the display unit 40 is also moved in the same direction (S32). [0063] When the left or right cursor key 56c or 56d is pressed to switch to the other cursor (S34), the switched value of the lower limit value or the upper limit value of the frequency range can be corrected, and the position of the ▽ mark 65 on the display 40 Also move to the other cursor side. In this state, when the shuttle key 58 is operated (S28), the corresponding value is changed in the direction of turning the shuttle key 58 (S29, S30, S31), and the マ ー ク mark 65 on the display 40 is also in the same direction accordingly. Move to (S32). [0064] Thus, when the frequency range is set and then the up key 56a or the down key 56b is pressed (S39), as shown in FIG. 15A, the level of the desired characteristic is pressed for the set frequency range. Depending on the number of times or pressing time, while maintaining 08-05-2019 21 continuity with the outside of the set frequency range, the central position of the frequency range is peaked and increased or decreased in a curve (S40, S41), the display unit 40 The display of the desired characteristics in the will also change accordingly. According to such a correction method, the operation is easy since it is only necessary to specify the frequency range and the amount of increase or decrease of the level. In addition, since the influence of level increase / decrease does not reach outside the designated frequency range, it is easy to grasp how the characteristic actually changes by the increase / decrease operation, and it is easy to correct it to the desired characteristic. The calculation for correcting the desired characteristic is performed by the calculation means 10 of FIG. [0065] As a specific correction processing algorithm in the calculation means 10, for example, it is examined according to the frequency range whether or not the correction of the operation sense and the actual characteristic match if the correction curve is increased or decreased according to the frequency range. The correction curve may be set in advance in a table, and the corresponding correction curve may be read out from the table according to the set frequency range, and a gain may be added and used according to the increase / decrease indication amount. By doing so, the sense of the level increase / decrease operation matches the actual change of the characteristic, and the correction operation to the desired desired characteristic becomes easy. [0066] If the up key 56a or the down key 56b is pressed with the frequency range lower limit value set to the lowest frequency of the entire frequency band, the desired characteristics are as shown in FIG. 15 (b). Change to the state of Similarly, if the up key 56a or the down key 56b is pressed with the frequency range upper limit value set to the highest frequency of all frequency bands, the desired characteristics are as shown in FIG. 15 (c). It changes to the state of going down. Also in these cases, for example, a correction curve of one side up or one side down corresponding to the frequency range and the increase / decrease amount is set in advance in the table, and the corresponding correction curve is read out from the table according to the set frequency range A gain corresponding to the amount (the number of times the up and down keys 56a and 56b are pressed) can be given and used. [0067] 08-05-2019 22 When the desired characteristic is corrected as described above, the key setting button 60 is pressed (S42) to exit the characteristic setting routine, and at this time, the characteristic determination and setting are completed (S43). Note that the determined characteristic can be stored and instructed as needed to be stored in the designated area of the backup power-supply RAM 138 as correction desired characteristic information and read out and used at any time. Therefore, it is not necessary to adjust again each time the desired characteristic is switched. [0068] Describes another modification method of modifying the desired characteristics. When the desired characteristics are set, it may be felt that the correction is excessive if the correction characteristics are obtained as it is and equalized. Therefore, as shown in FIG. 16, the intermediate characteristic between the initially set desired characteristic (the desired characteristic corrected as shown in FIG. 15 may be used) and the measurement characteristic is automatically calculated by the operation means 140. Can be newly set and used as a correction desired characteristic. Specifically, for example, the difference (that is, the correction value of each frequency) at each frequency between the initially set desired characteristic and the measured characteristic is equally divided into 20, and in accordance with this step, each time the up key 56a or the down key 56b is pressed The characteristic change is calculated and displayed so as to gradually bring the characteristic closer to the measurement characteristic or vice versa, and when the desired characteristic is obtained, this is set as a new desired characteristic. Do. FIG. 17 shows the calculation process at this time. First, the difference between the measured characteristic Nb and the initially set desired characteristic Db is determined (S51), and the difference Eb is multiplied by [the number of times the up key 56a or the down key 56b is pressed] / 20 to correct the desired characteristic ΔEb Is obtained (S52), this correction amount .DELTA.Eb is added to the desired characteristic Db to obtain Db + .DELTA.Eb (S53), and this is used as a new desired characteristic (S54). By doing this, it is possible to set an intermediate correction value that is not corrected as in the initial desired characteristic Db in a well-balanced manner by a simple operation in all frequency bands. Intermediate characteristics created in this manner can also be stored in the RAM 136. [0069] Calculation Characteristic of Correction Characteristic The correction characteristic of the correction characteristic is automatically calculated as the difference from the measurement characteristic by the calculation means 142 by setting the desired characteristic, and displayed on the display unit 40. 08-05-2019 23 [0070] Correction of correction characteristics If, for example, a desired characteristic of 0 dB flat is set with respect to the measurement characteristics shown in FIG. 18A, the correction characteristics have large peak dips as shown in FIG. 18B. This peak-dip often results from a slight change in the measurement environment, and when equalizing using such a correction characteristic as it is, a portion largely corrected (indicated by ○ in the same (b)) ), With slight changes in the environment (eg slight deviations in the frequency characteristics due to the influence of air temperature and humidity), so that the correction can no longer be a true correction and conversely usually not As shown in c), the correction error becomes large, resulting in rather peculiar characteristics. Therefore, the upper and lower limit values of the level of the correction characteristic are set to an arbitrary value (for example, ± 10 dB) by the operation of the operator. Thus, as shown in FIG. 18 (d), the correction characteristic calculation means 142 of FIG. 1 restricts the upper and lower limit values of the correction characteristic to the set value to prevent the correction from being performed more than necessary. Prevent the increase. Further, since the maximum value on the + side of the correction characteristic is limited by this, the maximum input can be suppressed, and distortion of the entire system such as the power amplifier and the speaker can be suppressed. [0071] Also, if the correction range is limited due to the limitation of the reproduction frequency characteristic of the speaker to be used, if the speaker is driven as it is based on the calculated correction characteristic, the speaker may be overloaded. The frequency range is set according to the above, and the correction characteristic is utilized only within that range, and the correction is not performed by making the outside of the range 0 dB flat. The frequency range to be corrected is displayed in the form of a horizontal bar graph as the corrected frequency range display 50 on the display unit 40 of FIG. [0072] As described above, FIG. 19 shows an example of a specific calculation process at each stage from the measurement data being obtained to the final determination of the correction 08-05-2019 24 characteristic. Data to be used for calculation of measurement characteristics is selected (S61) and weighted from among a plurality of times of measurement data stored in the band data memory 126 of FIG. 1 and selected. It is assumed that b = 1 to B by the band number obtained by dividing the selected data. B is 31 or 61 in this embodiment. [0073] When a plurality of data are selected, the group averaging means 130 obtains the group average for each band (S62). Then, an average value of all the bands of the group average is obtained (S63). Further, normalized average measurement data is obtained (S64), and this is displayed on the display unit 40 as a measurement characteristic. The measured characteristic Nb is splineinterpolated to be continuous data. By normalization, the average value of the measurement characteristics Nb is adjusted to be always 0 dB, and even if the sound collection level is small, the display of the measurement characteristics on the display unit 40 is always approximately on the same level. It becomes easy to compare with the characteristic display. [0074] When the desired characteristic Db is set by the operation of the operator (S65), Eb = Nb−Db is obtained as the correction characteristic in the calculation means 142 (S66). The measurement characteristic Nb here is data after spline interpolation. Then, the average value of all the bands of this correction characteristic is obtained (S67). Further, the calculating means 142 obtains a normalized correction characteristic (S68). By normalization, the average value of the correction characteristic Fb is adjusted to be always 0 dB, whereby the sound before and after correction as a whole is not changed in sound volume but only in sound quality. [0075] A process of restricting the upper limit value and the lower limit value of the level shown in FIG. 18D is performed on the obtained correction characteristic Fb (S69). In addition, the processes of steps S66 to S69 are performed only within the designated frequency range of FIG. 18 (e). Outside the designated frequency range, processing to flatten the correction characteristic by 0 dB is separately performed (S70). The correction characteristic finally determined in this manner goes to a routine for calculating a convolution (equalizer) filter coefficient (S71). 08-05-2019 25 [0076] Calculation of Equalizer Filter Coefficients The FIR filter algorithm for acoustic characteristic correction has advantages and disadvantages, and may not be usable depending on the purpose of use. Therefore, here, as the FIR filter, one of a linear phase filter and a minimum phase filter can be selected according to the selection operation of the operator. The impulse responses of the linear phase filter and the minimum phase filter are, for example, as shown in FIGS. 6 (d) and 6 (e), and their advantages and disadvantages are as follows. [0077] Transmission characteristics Delay amount Ease of calculation of filter coefficients Linear phase filter × × (large) ○ Minimum phase filter Δ ◎ (small) Δ According to this, the linear phase filter has good transmission characteristics and filter coefficient calculation is easy The delay is too large (see FIG. 6 (d)), and can not be used when real-time characteristics such as PA and mixdown are required (since the raw sound and the equalized sound deviate in time). In addition, although the minimum phase filter is inferior to the linear phase filter in terms of transmission characteristics and easiness of filter coefficient calculation, since there is almost no delay (see FIG. 6E), it is suitable when real time property is required. Therefore, one device can be used in various situations by allowing the operator to select one of the algorithms according to the purpose of use. [0078] In any case, since the correction characteristic assignment utilizes an FIR filter using digital convolution operation, it is possible to impart a linear phase filter, a minimum phase filter, or other special characteristics simply by switching the algorithm, The specification change is extremely easy, and if the calculation accuracy is arbitrarily increased as necessary, the correction accuracy can be set arbitrarily, and the practical effect of using the FIR correction means in this kind of acoustic characteristic correction device is large . [0079] An example of a procedure for calculating the impulse response of the linear phase filter and the impulse response of the minimum phase filter by using the inverse Fourier transform or the like from the correction characteristic in the equalizer filter coefficient calculation means 144 will be described. 08-05-2019 26 [0080] (A) Calculation of Impulse Response of Linear Phase Filter i) The correction characteristic is temporarily divided into bands (for example, every 1/3 to 1/12 octave pitch), and the power average for each band is determined. ii) The obtained power average value is used as a value at the center frequency of each band to interpolate into 4096-point data that can be Fourier-transformed by spline interpolation or the like. iii) Inverse Fourier transform is performed on complex-form data in which the data obtained in ii) is the real part (corresponding to the amplitude term) and the imaginary part (corresponding to the phase term) is all zero. iv) The real part of the resulting complex format data is directly used as a linear phase impulse response, so these are set as the coefficients of the FIR filter (convolution operator 34). [0081] (B) Calculation of Impulse Response of Minimum Phase Filter i) The correction characteristic is temporarily divided into bands (for example, every 1/3 to 1/12 octave pitch), and the power average for each band is determined. ii) The obtained power average value is used as a value at the center frequency of each band to interpolate into 4096-point data that can be Fouriertransformed by spline interpolation or the like. iii) The data obtained in ii) is used as the real part and the Hilbert transform is applied to the complex format data in which all imaginary parts are set to 0, and the complex format data that conforms to the correction characteristic curve and has a minimum phase shift Do. In this complex type data, necessary phase components are added to the imaginary part. iv) Inverse Fourier transform the complex format data obtained in iii). v) Since the real part of the resulting complex format data is the minimum phase impulse response, these are set as the coefficients of the FIR filter (convolution calculator 34). [0082] In addition to the linear phase filter and the minimum phase filter, a filter having an intermediate characteristic may be prepared, and an arbitrary filter may be selected therefrom. 08-05-2019 27 [0083] Confirmation of Correction Characteristics The coefficient of the FIR filter 34 is set in the abovedescribed procedure to confirm the correction effect, and the results are shown in FIG. (A) is an initial (i.e., without equalization) measurement result at each measurement point P1 to P5 (see FIG. 5). (B) shows a measurement characteristic in which measurement data of each of the points P1 to P5 are collectively averaged with the same weighting and a desired characteristic arbitrarily set by the operator. (C) is a correction characteristic obtained as a difference from the desired characteristic of (b). The FIR filter coefficients calculated based on the correction characteristics are set in the convolution unit 34 to configure an equalizer, and the measurement signal (band signal or TSP signal) is reproduced through the equalizer to perform measurement again. The result measured at each measurement point P1 to P5 is shown in FIG. According to this, it was confirmed that the corresponding characteristic correction was made at every measurement point and the optimum correction was made for the area including these points, as compared with that before the correction of (a). [0084] By inputting a music source instead of the music reproduction measurement signal and reproducing it through an equalizer (convolution calculator 34), it is possible to enjoy music appreciation with reproduction characteristics as desired. [0085] When a large number of speaker systems exist in a hall or the like, a correction device is required for each speaker system, but the music characteristic correction apparatus 10 with a response characteristic measuring function shown in FIGS. 1 and 2 is used for each system. The cost of equipment may be high. Therefore, in such a case, as shown in FIG. 21, an extension having only one system of acoustic characteristic correction apparatus 10 with a response characteristic measurement function and the other with an acoustic characteristic correction function without a response characteristic measurement function is prepared. Unit 11 can be used. In this case, when measuring the 08-05-2019 28 response characteristics for each of the systems SY1 to SYn, the measurement is carried out by sequentially changing the systems SY1 to SYn to this using the acoustic characteristic correction device 10 with the response characteristic measuring function, and the measurement results Are stored in the main unit 12 of the correction device 10, and the correction device 10 performs setting of desired characteristics of each system, calculation of correction characteristics, calculation of FIR filter (equalizer) coefficients, and communication of calculation results of FIR filter coefficients The cable 13 or the RAM card 148 is used to transfer to the expansion unit 11 of the corresponding system. Then, each expansion unit 11 can perform equalization according to the desired characteristic by setting the transferred FIR filter coefficient in the convolution unit 34. According to this, since the expansion unit 11 does not require the configuration for characteristic measurement and setting of desired characteristics, calculation of correction characteristics, correction, and calculation of FIR filter coefficients, the configuration can be simplified and equipment cost can be reduced. . [0086] [Modifications] In the above-described embodiment, in the calculation of the measurement characteristics, for example, spline interpolation is performed on the average value obtained for each divided band in any of the band signal method and the TSP method. Although the band division is performed again when calculating the FIR filter coefficient that realizes the correction characteristic as well as obtaining it, the present invention is not particularly limited to this. [0087] That is, when the obtained measurement characteristics are displayed with high accuracy, or when it is intended to use the measurement characteristics separately, it is difficult to use the divided band data as it is, but in other cases the correction characteristics are By performing spline interpolation at the time of calculation, it is possible to omit or simplify interpolation processing at the time of calculation of measurement characteristics before that. For example, the correction information is calculated as a correction value for each frequencydivided band based on the measured characteristics calculated for each frequency-divided band and the desired characteristics set for each frequency-divided band, and calculated for each band The correction value can also be calculated by interpolation as a value at about the center frequency of each band by interpolation to obtain a correction characteristic. In this way, it is possible to prevent the occurrence of a large peak-dip in the correction characteristics and to prevent the sense of incongruity due to extreme correction as a matter of course. The amount of calculation can be reduced in each stage as compared with the case and the like, and it is 08-05-2019 29 effective because the correction characteristic to be finally obtained does not significantly deteriorate in accuracy. [0088] As described above, according to the first aspect of the present invention, when the response characteristic is measured using the TSP signal as the measurement signal, the time compression by the inverse filter characteristic at the time of measurement and Since the common convolution operator is used to apply the correction characteristics, the hardware configuration is simplified and the apparatus can be miniaturized. [0089] Further, according to the invention of claim 2, as the number of stages of the convolution operation unit, the number of stages necessary for giving the correction characteristic is prepared, and the time compression by the inverse filter characteristic is divided in time, so that A small convolution unit can be used to share time compression and correction characteristics. 08-05-2019 30
© Copyright 2021 DropDoc