Patent Translate
Powered by EPO and Google
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.
Technical Field The present invention relates to a speech recognition apparatus. (2) In recent
years, development of speech recognition devices has been brisk, and specific speaker methods
and non-specific speaker methods have been put to practical use. However, the recognition rate
of this device also changes significantly if the use environment changes. For example, in the case
where a voice is picked up with a reflector such as a conference desk or a blackboard placed in
the vicinity of the microphone, the transmission frequency characteristic thereof is a comb-like
shape having a large number of dips, and the speaker-microphone-reflection Depending on the
physical relationship of the body, the level characteristics fluctuate greatly. This is one of the
recognition rate reduction factors when speech recognition is performed by DP matching or the
like. The recognition evaluation experiment using similar words is performed for the speech with
the distance between the speaker and the microphone for sound collection and with and without
the reflector, and the influence of the transmission frequency characteristics on the recognition
performance The Proceedings of the Acoustical Society Conference, March 1986, 269.270.
According to the report, three microphones arranged at different distances from the speaker's
mouth were used simultaneously to measure the height from the reflector to the sound source
(speaker) and the height from the reflector to the microphone. When the distance from the sound
source to the microphone is 10.50, 90], there are no reflectors and there is no intra-individual
variation of the speaker (fluctuation of the speech every time the user speaks) The recognition
rate in the case is almost the same due to the difference in microphone pickup distance. ■ The
recognition rate in the case of intra-individual fluctuation has caused about 20% fluctuation. (1)
The recognition rate in the presence of a reflector decreases as the distance between the speaker
and the microphone for sound collection increases, and is particularly remarkable in the case of L
= 90 an. In addition, the value changes with speakers. (2) Even when there is a reflector, the
recognition rate on the L = 50 side is a range included in the amount of personal white variation
of the speaker, but is reduced by several percent compared to the case of L = 10− under the
same conditions. It is believed that he understood. Thus, there is a difference in recognition rate
between when there is a reflector nearby and when it is not. This is noticeable, for example, when
using a speaker-independent recognition system in a car. In the specific speaker system, although
it is possible to avoid to some extent by creating a standard pattern in the use environment, it is
not possible to formulate measures against this because the specific speaker does not know what
kind of environment it will be used in . That is, in a narrow space such as in a car, the frequency
characteristics of the microphones that reflect sound are often changed to lower the recognition
The present invention was made in view of the actual situation such as a child ream, and in
particular, it was made for the purpose of providing a speech recognition apparatus which can
correctly recognize even in a place where acoustic characteristics are different. . In order to
achieve the above object, the present invention converts a sound into an electric signal. ! ! ! A
speech recognition apparatus comprising an electrical converter, an analysis unit for analyzing
the same, and a pattern comparison unit for comparing the analyzed results, comprising: an
oscillator for emitting an acoustic signal of an audio frequency band; It is characterized in that
No. B is analyzed, and the analysis result is corrected by the analysis unit or the route before and
after it so as to become a defined relationship. Hereinafter, the present invention will be
described based on embodiments of the present invention. FIG. 1 is a block diagram for
explaining an embodiment of the present invention, in which 1 is an oscillator, 2 is an amplifier,
3 is a speaker, 4 is a microphone, 5 is a microphone amplifier, 6 is a band pass filter group , 7 is
an attenuator, 8 is a register, 9 is an average / difference operation unit, 10 is a comparison unit,
11 is a dictionary, and a microphone 4 which is an electroacoustic transducer and a signal
amplified by the microphone amplifier 5 which amplifies its electric signal Is input to a filter
panchromatic which is an analysis unit and subjected to frequency analysis. Although the
analysis by the filter bank is described as one example here, it is not limited to the filter bank but
any other method such as LPC may be used. The filter panchromatic is an array of several band
pass filters. The attenuator 7 is included in the output of the filter panchromatic so that the level
of each filter output can be adjusted. Initially, all attenuators are set so that all outputs are at the
same level. In a normal case, the analysis result placed in the register 8 and the standard pattern
of the dictionary 11 are compared by the comparison unit 10. Although the method of
comparison is not particularly limited, there is a typical one using dynamic programming (DP
matching). The process of recognition is omitted here as it does not matter here. Since the
frequency characteristic of the microphone is disturbed by the influence of ambient reflection as
it is, the method of the correction will be described below. The sine wave of the voice band (about
150 to 10 kllz) is swept from the oscillator 1 and the speaker 3 is driven L 2 to make a sound,
which is received by the microphone 4. At this time, if possible, for example, as shown in FIG. 2, it
is preferable to place the microphone 4 in a normal position and place the speaker 3 near the
lips. After the entire band has been swept, the outputs of the filters stored in the register 8 are
compared to obtain the difference between the values of adjacent registers, and the attenuator 8
is moved so as to reduce the difference.
More specifically, FIG. 3 (a) plots the output of the oscillator 1 before correction, and the O mark
in the drawing indicates the frequency position of the band pass filter. Here, the number of filters
is 8 ch. Next, the average value is determined from the eight values shown in FIG. This is shown
by the broken line in FIG. (A) The average / difference calculation unit 9 determines how much
the eight values in the figure are weaned from the average value, thereby moving the attenuator
7. The moving method is determined so that the correction value yL of the attenuator of each
band pass filter is proportional to y L = X Xt, where the average value is Ma and the output of
each band pass filter is xt (i = 1 to 8) There is. That is, it becomes a circle in FIG. 4 (b), and this
correction corrects the frequency characteristic as shown in FIG. 4 as shown in FIG. Although not
shown in FIG. 1, since the output of the filter is sampled at about 10 m5, it is considered to be
time-averaged within the sample period, and it is not necessary to consider the phase change of
the filter. . According to the present invention, as is apparent from the above description, it is
possible to return the frequency characteristics of the microphone to be flat and to prevent a
decrease in the recognition rate.
Brief description of the drawings
1 is a block diagram for explaining an embodiment of the present invention, FIG. 2 is a block
diagram for explaining an example of installation positions of a speaker and a microphone, and
FIGS. 3 and 4 are It is a figure for demonstrating the operating characteristic of this invention.
Reference Signs List 1 oscillator 2 amplifier 3 speaker 4 microphone 5 microphone amplifier 6
band pass filter group 7 attenuator 8 register. 9 ... average difference calculation unit, 10 ...
comparison unit, 11 ... dictionary. Figure 1 Figure 2 Figure 4