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 JP2004015125 An automatic gain control device capable of realizing excellent characteristics with respect to a predetermined input audio signal. SOLUTION: A voice signal inputted to a voice input terminal 1 is divided by a plurality of band division filter means 3 into signals of different frequency bands preset for each band division filter means 3, and each divided signal is divided The absolute value conversion means 4 converts it to an absolute value, the multiplication means 6 multiplies it with the weighting data stored in the weighting data storage means 5, and all the multiplied values are added up by the addition means 7. This value is corrected by the gain correction means 8 and compared with the threshold level data stored in the threshold level storage means 9 by the comparison means 10, and the gain generation means 11 generates a gain according to the comparison result. The radiation value generating means 12 makes a gain value having a time constant, and the output signal multiplying means 13 multiplies it with the input voice signal and outputs it. [Selected figure] Figure 1 Auto gain control device BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an automatic gain control device which makes the level of an output signal constant regardless of the level of an audio input. 2. Description of the Related Art Conventionally, this type of auto gain control device is shown as shown in FIG. In this automatic gain control device, the audio signal inputted from the audio input terminal 101 is converted into an absolute value by the absolute value conversion means 103, normalized by the gain correction means 104 to a predetermined gain, and corrected. The corrected gain is compared with the threshold level coefficient stored in advance in the threshold level storage unit 105 by the comparison unit 106. The gain generation means 107 generates a gain smaller than 1 when the gain is larger than the threshold level, and the output level as an output signal is attenuated. When the gain is smaller than the threshold 08-05-2019 1 level, the gain generation means 107 generates a gain larger than 1 and the output level as an output signal is amplified. In order to operate unnaturally with respect to a sudden change in the input signal, the radiation generation means 108 moderates the amount of change in the gain value generated by the gain generation means 107 and changes the gain. Eliminate the unnaturalness. The output multiplying means 109 multiplies the gain value having a time constant output from the wire generating means 108 by the audio signal data input from the audio input terminal 101, and outputs the output value. The signal is output from the audio output terminal 102 as a signal. In this way, it is possible to maintain a constant output signal level near the threshold level. However, in such a conventional auto gain control apparatus, since the frequency characteristic of the input audio signal is not taken into consideration, the auto gain control apparatus can not perform automatic operation depending on the frequency characteristic of the input audio signal. There is a problem that the operation of the gain control device varies. For example, even when noise having a high frequency component is erroneously input, the gain may be controlled as in a normal audio signal. Low-range noise such as background noise also causes malfunction. In addition, in the audio signal of the announcement and the music source and the like of the announcement where the frequency characteristic completely differs, the level where auto gain control operates becomes different, operation of auto gain control depends on the source of input audio signal There was also a problem that The present invention has been made to solve such a problem, and provides an automatic gain control device capable of realizing good characteristics with respect to a predetermined input audio signal. According to the present invention, there is provided an auto gain control apparatus comprising: band division filter means for dividing an input voice signal into a plurality of different frequency bands; and each of the plurality of different frequency bands. Means for converting the absolute value into absolute value, weighting data storage means for storing weighting data for each value absolute value converted by the absolute value conversion means, and each value converted by the absolute value conversion means A multiplication means for multiplying a value and weighting data stored in the weighting data storage means, and an addition means for adding the respective values multiplied by the multiplication means, and a gain based on the output of the addition means Control configuration. According to this configuration, the input speech signal is weighted for each predetermined frequency band, and the automatic gain control is performed based on the weighted signal. Here, weighting data is stored in the weighting data storage unit, and the weighting data switching unit is configured to switch the weighting data to be output to the multiplying unit among the plurality of sets of weighting data. . According to this configuration, the weighting for each frequency band of the input speech signal can be switched. Further, a hold control means for maintaining a gain generated for a predetermined time when the fluctuation amount of the output of the addition means is equal to or more than a preset value, and a time for maintaining the gain by the hold control means A plurality of hold time data storage means for storing the plurality of hold time data storage means, wherein the hold time data storage means sets, in the hold control means, a 08-05-2019 2 time for maintaining the gain corresponding to the weighting data switched by the weighting data switching means; did. According to this configuration, the maintenance time of the generated gain is also switched corresponding to the weighting for each frequency band of the switched input audio signal. Further, the image processing apparatus further comprises conversion processing means for converting an input voice signal into frequency data, and the weighting data switching means is configured to switch to weighting data corresponding to the frequency data. According to this configuration, the weighting can be automatically switched to the weighting suitable for the frequency characteristic of the input audio signal. BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. 1 to 4 show an auto gain control apparatus according to an embodiment of the present invention. As shown in FIG. 1, in the automatic gain control device of this embodiment, an audio input terminal 1 to which an audio signal is input, an audio output terminal 2 to which an audio signal whose gain is controlled is output, and A plurality of band division filter means 3 for passing only signals of different frequency bands that have been set, an absolute value conversion means 4 for performing absolute value conversion of the signals output from each of the band division filter means 3, and a band division filter means 3 Weighting data storage means 5 for storing data to be weighted for each divided frequency band, multiplication means 6 for multiplying weighted data by the value converted into absolute value, and addition means 7 for adding weighted values Gain correction means 8 for correcting the added value, and a threshold for storing threshold level data to be compared with the corrected value Field level storage means 9, comparison means 10 for comparing the corrected value with threshold level data, gain generation means 11 for generating a gain value according to the comparison result of comparison means 10 A winding wire generation means 12 and an output signal multiplication means 13 for multiplying the input speech signal by the gain value outputted by the radiation wire generation means 12 are provided. In such an automatic gain control apparatus, the audio signal input to the audio input terminal 1 is input to each of the plurality of band division filter means 3. Each band division filter means 3 outputs only signals of different frequency bands preset for each band division filter means 3. For example, when the frequency characteristic of the assumed input source is as shown in FIG. 2, four band-splitting filter means 3 having frequency characteristics as shown in FIGS. 3 (a) to 3 (d) are used. Do. The addition characteristic of these four band division filter means 3 is as shown in FIG. 4, and this addition characteristic approximates the characteristic of the assumed input source of FIG. If the number of divisions by the band division filter means 3 is increased, the input source can be closely approximated, and good gain control can be performed in accordance with the input source. The output signals of such band division filter means 3 are input to the absolute value conversion means 4 and subjected to absolute value conversion for each band. Weighting data storage means 5 stores weighting data for each value obtained by absolute value 08-05-2019 3 conversion of the signal output from each band division filter means 3. The weighting data reflects the characteristics of the frequency band of the input source assumed, and the weighting data of the frequency band serving as the center of gain control is set to a larger value. For example, when the four band division filter means 3 of the frequency characteristic as shown in FIG. 3 described above are used, the third filter output (the output of the band division filter means 3 having the characteristic of FIG. If the weighting for iii) is increased, the contribution of the third filter output band in the gain control (the frequency band divided by the band dividing filter means 3 having the characteristic of FIG. 3C) is increased. The weighting data for each frequency band divided by the band division filter means 3 stored in the weighting data storage means 5 is the value of the output of the corresponding band division filter means 3 converted into an absolute value. Then, all the values multiplied by the multiplication means 6 and multiplied by the multiplication means are added by the addition means 7. Thus, the signal output from the addition means 7 is a signal weighted by weighting data based on voice band model data corresponding to the predetermined band division filter means 3 stored in the weighting data storage means 5 in advance. Become. The output (gain value) of the addition means 7 approaches a predetermined gain value when an audio signal having the frequency characteristic of the audio band model data is input. The gain correction means 8 corrects by multiplying the output of the addition means 7 by a preset coefficient. Since the gain correction means 8 can adjust the absolute value of the gain value, it is only necessary to consider only the relative value for each frequency band as the weighting data. The gain value corrected by the gain correction means 8 is compared with the threshold level data stored in the threshold level storage means 9 in the comparison means 10. The gain generation means 11 generates a gain value smaller than a preset value 1 when the gain value is larger than the threshold level data. Thus, the output level of the output signal is attenuated and approaches the threshold level. When the gain value is smaller than the threshold level, the gain generation means 11 generates a gain value larger than a preset one. Therefore, the output level of the output signal is amplified and approaches the threshold level. When the weight of the output of a specific band dividing filter unit 3 is increased, when a sound source including a large number of frequency components of the frequency band divided by the band dividing filter unit 3 is input, the adding unit according to the input signal is added. The output value of 7 becomes large, it is easy to exceed the threshold level, and it is possible to obtain the operating characteristic of the auto gain control linked to a specific frequency band. The gain value generated by the gain generation means 11 is input to the radiation line generation means 12 and controlled so as to have an attack time and a release time having a predetermined time constant. The attack time in this case is a time constant for changing the gain value when the gain value generated by the gain generation means 11 decreases, and the release time is when the gain value generated by the gain generation means 11 increases. The time constant for changing the gain value of By setting the attack time and the release time, it is possible to set the response to the rapid level change (whether or not the rapid level change is to 08-05-2019 4 be followed). The output signal multiplication means 13 multiplies the gain value having a time constant by the radiation line generation means 12 by the audio signal inputted from the audio input terminal 1, and the output signal multiplication means 13 The output signal is the output signal of the auto gain controller. As described above, in the present embodiment, since the frequency characteristic of the predetermined voice is divided and approximated by the band division filter means 3, it is possible to perform the automatic gain control aiming at the predetermined voice signal. Further, since weighting is performed separately for each band divided by the band division filter means 3, even if an audio signal of an unexpected band is input, the weighting coefficient of that band is small, so the change of the gain value is As a result, the variation of the gain value by the gain generation means 11 is eliminated, and the variation of the level of the signal outputted from the audio output terminal 2 can be eliminated. In addition, because it is possible to emphasize only the characteristic frequency band for each audio source by the amount of weighting by the weighting factor, for audio sources having other frequency characteristics, a fixed gain is used. It can be maintained and can react only to an audio source having a targeted frequency characteristic. As a first mode of the present embodiment, as shown in FIG. 5, weighting data for each divided band according to several types of audio sources is stored in the weighting data storage means 5. For example, as the frequency characteristics are different according to the average distribution of the frequency characteristics of the male announcement and the female announcement, weighting data by voice band is prepared according to the application and the feature to be announced. This may be measured beforehand and summarized statistically, and the distribution of the average frequency characteristic may be used as weighted data. The weighting data switching unit 14 switches the weighting data output from the weighting data storage unit 5 to the multiplying unit 6 by the operation of the user. The weighting data switching means 14 may be a physical switch structure or a switch on a GUI (Graphical User Interface) of a personal computer. For example, based on the frequency characteristics of a male announcement and a female announcement, weighting data corresponding to each is stored in the weighting data storage means 5, and an auto gain control device is used for a voice signal of a male announcement. In this case, the weighting data switching unit 14 outputs weighting data based on the frequency characteristic of the male announcement to the multiplying unit 6. Further, when using an automatic gain control device for a female announcement voice signal, the weighting data switching means 14 switches to output weighting data based on the frequency characteristic of the female announcement to the multiplication means 6. Alternatively, data of frequency characteristics of a specific person may be measured in advance, and the data may be stored in the weighting data storage unit 5 as a data model classified by voice band. By configuring in this way, it is possible to select weighting data that matches the audio source by the weighting data switching means 14, and it is possible to perform automatic gain control that corresponds to the frequency characteristic unique to the audio source. As a second other aspect of the present embodiment, as shown in FIG. 6, each of the weighting data storage means 5 08-05-2019 5 includes, for each of several types of audio sources, as in the first other aspect described above. A weighting data switching unit 14 is provided which stores weighting data for each divided band, and switches the weighting data that the weighting data storage unit 5 outputs to the multiplying unit 6. The hold control circuit 15 holds the gain of the gain generation means 11 for a certain period of time according to the judgment result of the comparison means 10, and a plurality of hold time data (hold time data) of the hold control circuit 15 are stored. A hold time data storage means 16 is provided which selects one of the hold time data stored in accordance with the switching instruction of the switching means 14 and outputs the selected one to the hold control circuit. In such an automatic gain control apparatus, when the user operates the weighting data switching means 14 to select one weighting data, the weighting data storage means 5 outputs the selected weighting data to the multiplying means 6 . Further, the hold time data storage unit 16 outputs hold time data corresponding to the weighting data selected by the weighting data switching unit 14 to the hold control circuit 15. The comparison means 10 compares the gain value corrected by the gain correction means 8 with the threshold level data stored in the threshold level storage means 9 and changes in the difference between the gain value and the threshold level data When the amount exceeds a preset value, a trigger is output to the hold control circuit 15. When the trigger is input, the hold control circuit 15 starts an internal timer and outputs the gain value generated by the gain generation means 11 only for the time of the hold time data which is output and set by the hold time data storage means 16. To last. With this configuration, the hold time of the gain can also be switched according to the audio source in conjunction with the switching of the weighting data, and the unnatural gain fluctuation of the output level can be suppressed for each audio input source. Can. For example, in the case of an audio source with a large amount of gain fluctuation, hold time data storage corresponding to the weighting data applied to the audio source, when the fluctuation of the output level is prevented from becoming unnatural as much as possible. It is preferable to store a longer time as hold time data stored in the means 16. Further, when gain variation of the audio source can not be estimated, and the response response of the output level is desired to be fast, the hold is stored in the hold time data storage means 16 corresponding to the weighting data applied to the audio source. It is good to store short time as time data. The hold time data may be determined so as to make optimum gain fluctuation for each audio source based on data measured in advance. As a third other aspect of the present embodiment, as shown in FIG. 7, each of the weighting data storage means 5 for each of several types of audio sources is the same as the above-mentioned second other aspect. Hold control circuit 15 which stores weighting data for each divided band and holds the gain of gain generation means 11 for a certain period of time according to the determination result of comparison means 10, selects one of the stored hold time data and holds it. And a hold time data storage unit configured to set an output to the control circuit. 08-05-2019 6 Also, a fast Fourier transform (FFT) processing means 17 that performs fast Fourier transform (FFT) processing on the voice signal input to the voice input terminal 1 and analyzes the frequency characteristics of the input voice signal. And data model judging means 18 for selecting the closest model data from several kinds of voice band classified model data registered based on the analysis result of the fast Fourier transform processing means 17. In such an automatic gain control apparatus, the FFT processing means 17 performs FFT processing on the audio signal input to the audio input terminal 1 to analyze the frequency characteristics of the input audio signal, and analyze the analysis result information as data. It outputs to the model judgment means 18. The data model judging means 18 selects, based on the analysis result information of the FFT processing means 17, voice band model data having frequency characteristics close to the frequency characteristics of the input voice signal, and selecting the selected voice band Weighting data corresponding to model data is selected from the weighting data storage means 5. The weighting data storage unit 5 outputs the selected weighting data to the multiplication unit 6. Further, the hold time data storage unit 16 outputs hold time data corresponding to the weighting data selected by the data model determination unit 18 to the hold control circuit 15. As in the second other aspect described above, the comparator 10 triggers the hold control circuit 15 when the amount of change in the difference between the gain value and the threshold level data becomes equal to or greater than a preset value. Output When a trigger is input, the hold control circuit 15 maintains the gain value generated by the gain generation means 11 for the hold time data time set by the hold time data storage means 16. When the FFT processing is performed by a DSP (Digital Signal Processor) or the like, a certain amount of processing time is required, so the FFT processing is performed at time intervals for this processing time. By configuring in this manner, weighting data and hold time data corresponding to the voice band classified data model that automatically matches the frequency characteristics of the input voice signal are automatically selected, and auto gain control is matched to the input voice signal. Can be done automatically. As described above, according to the present invention, the input audio signal is weighted for each frequency band set in advance, and automatic gain control is performed based on the weighted signal. It is possible to operate only on the input speech signal having the frequency characteristic of and to follow the assumed frequency characteristic of the speech signal. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a schematic block diagram showing an automatic gain control apparatus according to an embodiment of the present invention. FIG. 2 is a schematic diagram of an input source assumed for the automatic gain control apparatus according to an embodiment of the present invention. FIG. 3 is a diagram showing an example of frequency characteristics. FIG. 3 is a diagram showing an example of frequency characteristics of each band division filter means of an automatic gain control apparatus according to an embodiment of the present invention. FIG. 5 is a schematic block diagram showing an auto gain control device according to a first other aspect of the embodiment of the present invention. FIG. 7 is a schematic block diagram showing an auto gain control apparatus according to a second 08-05-2019 7 other aspect of the embodiment of the present invention. FIG. 7 is an auto gain controller according to a third other aspect of the embodiment of the present invention 8 is a schematic block diagram showing a conventional auto gain control device. [Explanation of symbols] 1 voice input terminal 2 voice output terminal 3 band division filter means 4 absolute value conversion means 5 weighting data storage Means 6 Multiplication means 7 Addition means 8 Gain correction means 9 Threshold level storage means 10 Comparison means 11 Gain generation means 12 Radiation line generation means 13 Output signal multiplication means 14 Weighted data switching means 15 Hold control circuit 16 Hold time data storage means 17 Fast Fourier Transform Processing Means 18 Data Model Judgment Means 101 Voice Input Terminal 102 Voice Output Terminal 103 Absolute Value Converting Means 104 Gain Correction Means 105 Threshold Level Storage Means 106 Comparison Means 107 Gain Generation Means 108 Irregular Wire Generation Means 08-05-2019 8
© Copyright 2021 DropDoc