JP2002156997A - Voice detection controller - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a voice detection controller which can reduce the power consumption of a voice processor by performing voice detection adaptively to environment. SOLUTION: An input signal is converted by a rectifier 101 into an amplitude signal, a threshold is found by a low-pass filter 102 and an amplifier 103, and the quantity of variation in amplitude is found by a band-pass filter 104 and a rectifier 105. A comparator 106 compares the found variation quantity with the threshold and judges that the input signal possibly includes a voice when the variation quantity exceeds the threshold. Then a timer 107 outputs a control signal for a specific time and the voice processor is placed in operation.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice detection control device used for a voice processing device such as a voice input device, a voice detection device or a voice recognition device.

[0002]

2. Description of the Related Art Digital voice processing devices represented by voice recognition devices generally include DSPs and CPUs capable of high-speed processing.
And requires large power during operation. For this reason, it is necessary to always use a commercial power supply for processing the input signal or stop the operation until a switch operation is performed.

In order to improve this, a low power consumption sound volume detection circuit is provided, and only this sound volume detection circuit is constantly operated, and when the sound volume becomes larger than a predetermined value, an AD (analog-digital) conversion device is provided. Some proposals have been made to start the operation of the speech recognition device. This volume detection circuit is realized by an analog circuit or a simple CMOS circuit. For example, Japanese Patent Application Laid-Open No. 2000-89792 proposes a speech recognition apparatus that performs speech detection using a CMOS logic circuit.

[0004]

When a human has a conversation, he or she unconsciously adjusts the volume of the voice in accordance with the environmental noise. Therefore, in a quiet environment, the user usually speaks with a small voice, and in a noisy environment, the voice is loud. Put out. For this reason, when a certain threshold value is provided for the amplitude of the input signal in the volume detection circuit, the voice recognition device does not operate because a small voice cannot be detected in a quiet environment, and conversely, Below, the operation of the speech recognition apparatus assuming that a voice is always included results in wasteful power consumption.

When the recognition device is operated after detecting the sound volume, the voice signal sent to the voice recognition device lacks the beginning of a word, which is not desirable.

Further, when the background noise is large, the amplitude of the voice may be smaller than the amplitude of the noise, so it is difficult to determine the presence or absence of the voice only from the amplitude of the input signal or the short-time energy. In some cases, the beginning or end of a word may be missing or detection may not be possible.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems of the prior art, and performs voice detection adapted to an environment to reduce power consumption of a voice processing apparatus. It is intended to provide a device.

[0008]

A voice detection control device according to the present invention comprises a rectifier for rectifying an input signal and converting it into an amplitude signal, and individually processing the amplitude signal from the rectifier to change the amplitude. By comparing the output from each filter means with two types of filter means having different characteristics for obtaining the amount and the threshold value, respectively,
A comparison unit that outputs a voice detection signal by determining that the input signal may include voice when the amplitude change amount exceeds a threshold, and outputs a voice detection signal from the comparison unit. A timer means for outputting a control signal for a predetermined time when the operation is performed, thereby achieving the above object.

A voice detection control device according to the present invention comprises: a rectifier for rectifying an input signal and converting the input signal into an amplitude signal; a logarithmic amplifier for logarithmically converting the input signal converted to an amplitude signal by the rectifier; A band-pass filter for processing the output from the amplifying unit, and determining that there is a possibility that voice is included in the input signal when the output from the band-pass filter exceeds a predetermined threshold. And a timer means for outputting a control signal for a predetermined time when the sound detection signal is output from the comparing means, thereby achieving the above object. You.

A second rectifier for re-rectifying the amplitude signal processed by the band-pass filter, wherein an output from the band-pass filter is input to the comparator via the second rectifier; Good.

Amplifying means for amplifying the amplitude signal processed by the low-pass filter means, wherein an amplification factor of the amplifying means can be adjusted from the outside, and an output from the low-pass filter means is passed through the amplifying means. May be input to the comparing means.

An analog-to-digital converter for converting the input signal into digital data, and a buffer memory;
The input signal before voice detection may be stored as digital data in the buffer memory.

A plurality of amplifying means for amplifying the output from the low-pass filter means, a plurality of comparing means, and a plurality of timer means are provided, and a plurality of control signals are generated by setting the amplification factors of the respective amplifying means to different values. One of them may control a power supply and a clock of the analog-to-digital converter and the buffer memory.

A plurality of second band-pass filter means for separating the input signal for each frequency band, and a plurality of the plurality of band signals for converting each band signal separated by each second band-pass filter means into an amplitude signal. Rectifier means, a plurality of sets of two types of filter means having different characteristics, each of which individually processes an amplitude signal from the rectifier means to obtain an amplitude change amount and a threshold value, and a filter means for each frequency band. A plurality of the comparing means for outputting a sound detection signal by judging that there is a possibility that sound is included in the input signal when the amount of change in amplitude exceeds a threshold value. And a totaling means for synthesizing the voice detection signal from each comparing means, and the timer means for outputting a control signal for a predetermined time based on an output from the generalizing means.

A plurality of second band-pass filter means for separating an input signal for each frequency band, and a plurality of the plurality of band signals for converting each band signal separated by each second band-pass filter means into an amplitude signal. Rectifier, a plurality of logarithmic amplifiers that logarithmically convert the amplitude signal for each frequency band from the rectifier, and a plurality of the bandpass filter units that process the output from the logarithmic amplifier for each frequency band. When the output from the band-pass filter exceeds a predetermined threshold, it is determined that there is a possibility that voice is included in the input signal of the frequency band, and a plurality of voice detection signals are output. Even if it is provided with a comparing means, a general means for synthesizing the voice detection signal from each comparing means, and the timer means for outputting a control signal for a predetermined time by an output from the general means. There.

A second low-pass filter for processing an amplitude signal for each frequency band converted by the second band-pass filter and the rectifier, and an output from the second low-pass filter; A multiplexer and an analog-to-digital converter for converting
A buffer memory may be provided, and an amplitude signal for each frequency band may be stored as digital data in the buffer memory.

An average of the amplitude signals for each frequency band converted by the second band-pass filter means and the rectifier means by the low-pass filter means,
After reducing the influence of noise by subtracting from the amplitude signal for each frequency band, the data may be converted into digital data.

A voice detection control device according to the present invention comprises a plurality of amplitude modulation means for amplitude-modulating an input signal at a plurality of different frequencies, and a plurality of amplitude modulation means for processing an output from each amplitude modulation means, each having a predetermined characteristic. 3 low-pass filter means, and further individually processes the amplitude signal for each frequency band converted by the amplitude modulation means and the third low-pass filter means, thereby obtaining an amplitude change amount and a threshold value. A plurality of sets of two types of filter means having different characteristics are respectively obtained, and outputs from the respective filter means are compared for each frequency band, and when the amount of change in amplitude exceeds a threshold value, A plurality of comparing means for determining that there is a possibility that sound is included and outputting a sound detection signal, a total means for summing the sound detection signals from the respective comparing means, and a predetermined time of, And a timer means for outputting a control signal, the object is achieved.

A voice detection control device according to the present invention comprises a plurality of amplitude modulation means for amplitude modulating an input signal at a plurality of different frequencies, and a plurality of amplitude modulation means for processing an output from each amplitude modulation means, each having a predetermined characteristic. And a plurality of logarithmic amplifiers that logarithmically convert the amplitude signal for each frequency band from the rectifier, the amplitude modulator and the third low-pass filter. A plurality of logarithmic amplifiers that logarithmically convert the converted amplitude signal for each frequency band, a plurality of bandpass filter units for processing the output from the logarithmic amplifier unit for each frequency band, If the output exceeds a predetermined threshold,
A plurality of comparing means for judging that there is a possibility that sound is included in the input signal of the frequency band and outputting a sound detection signal; a total means for summing the sound detection signals from the respective comparing means; For a predetermined time,
Timer means for outputting a control signal, whereby the above object is achieved.

[0020] An analog-to-digital converter and a digital processor may be provided, and the input signal may be digitally processed by software.

Hereinafter, the operation of the present invention will be described.

In the present invention, the input signal is rectified by the rectifier and converted into an amplitude signal, and the amplitude signal is individually processed by two types of filter having different characteristics. Specifically, the number H of the amplitude signals is determined by the low-pass filter means.
An average amplitude of an input signal is obtained by extracting only low-frequency components equal to or lower than z. The average amplitude is amplified by the amplification means at a predetermined magnification, and is output as a threshold value for the change amount of the amplitude. The threshold can be adjusted by externally adjusting the amplification factor (gain) of the amplifying means.

Further, a change component of several Hz to about 20 Hz contained in the amplitude signal of the input signal is obtained by the band-pass filter means. By rectifying this change component again by the second rectifier, it is possible to express a positive value whether the change in the amplitude is in the increasing direction or in the decreasing direction.

Then, a signal indicating the amount of change in amplitude is compared with the threshold value by a comparator, and if the amount of change in amplitude exceeds the threshold value, it is determined that there is a possibility that voice is included in the input signal. Judge and output a voice detection signal. When the sound detection signal is output from the comparing means, the control signal is output for a predetermined time by the timer means.

According to another aspect of the present invention, the input signal is rectified by the rectifying means, converted into an amplitude signal, and logarithmically amplified by the logarithmic amplification means. The logarithmic amplitude signal is processed by band-pass filter means to obtain a logarithmic amplitude change component. Also in this case, by rectifying the change component again by the second rectifier, the change in the amplitude can be expressed as a positive value regardless of whether the change is in the increasing direction or the decreasing direction.

Then, a signal indicating the amount of change in amplitude logarithmized by the comparator is compared with a predetermined threshold value. When the signal indicating the amount of change in logarithmized amplitude exceeds the predetermined threshold value, It determines that there is a possibility that the signal contains sound, and outputs a sound detection signal. When the sound detection signal is output from the comparing means, the control signal is output for a predetermined time by the timer means.

Further, the input signal may be converted into digital data by an analog-to-digital converter, and the input signal before voice detection may be stored as digital data in a buffer memory. At this time, a plurality of control signals are generated by providing a plurality of amplifying means, comparing means, and timer means for amplifying the output from the low-pass filter means and setting the amplification factors of the respective amplifying means to different values. Thus, it is possible to control the power supply and clock of the analog-to-digital converter and the buffer memory at a timing earlier than the voice detection.

Further, it is possible to provide a filter bank by providing a plurality of second band-pass filter means, to separate the input signal for each frequency band, and to detect the sound of each signal for each band. In this case, it is possible to integrate the voice detection signals from the respective comparing means by the integrating means, and to output the control signal from the timer means by the output from the integrating means.

Further, second low-pass filter means for processing an amplitude signal for each frequency band is provided. The output is time-divided by a multiplexer and input to an analog-to-digital converter. Can be stored in the buffer memory as digital data.

Furthermore, the influence of noise can be reduced by averaging the amplitude signal for each frequency band by the low-pass filter means, subtracting the average from the amplitude signal for each frequency band, and converting it to digital data. It is possible.

Alternatively, the amplitude signal is amplitude-modulated at a plurality of different frequencies by the amplitude modulation means, and then processed by a plurality of third low-pass filter means each having a predetermined characteristic, whereby the amplitude signal for each frequency band is obtained. May be converted to

These processes can also be digitally processed by software using an analog-to-digital converter and a digital processor.

[0033]

Embodiments of the present invention will be described below.

The amplitude of the environmental noise changes by a factor of 1000 (60 dB) or more depending on the situation, while the amount of human perception of the sound pressure is generally expressed in a logarithmic region, so that a considerably large environmental noise or a fairly small environmental noise is obtained. Unless otherwise, the speaker speaks out several to several hundred times as loud as the environmental noise. Further, in the voice utterance section, the steady section of the amplitude is short and changes frequently. Therefore, in a speech utterance section in which a person can sufficiently hear the sound, an amplitude change that is several times as large as the lower limit of the sound pressure of the environmental noise is observed.

Therefore, in the present invention, in order to detect voice, (1) a method of comparing the amount of change in amplitude with a threshold value adapted to the amplitude of environmental noise, or (2) a logarithmic amplitude Are compared with a fixed threshold value. These methods allow voice detection for various environmental noise levels.

In the above-mentioned method (1), the amplitude signal of the input signal is processed by a plurality of filters having different characteristics.
The sound is detected by comparing these. Specifically, the amplitude level of the environmental noise is obtained using a low-pass filter of several Hz or less, and the amplitude level is amplified or attenuated by an amplifier, thereby obtaining a threshold value adapted to the environment. The amplification or attenuation of the amplitude level is performed according to the characteristics of the input microphone and the characteristics of the filter, and is adjusted according to the application. The signal is slightly attenuated for easy detection, and slightly amplified for difficult detection. Note that a filter referred to as a low-pass filter in this specification can be realized by a band-pass filter that cuts off direct current and passes only low frequencies in order to eliminate the influence of direct current.

Further, a change component of amplitude due to voice is captured by using a band-pass filter. In this case, in order to remove a DC component or a fast changing component due to noise other than voice, it is desirable to set the low cutoff frequency of the bandpass filter to several Hz or less and set the high cutoff frequency to several Hz to about 20 Hz. .

In the above method (2), the amplitude signal of the input signal is logarithmically amplified and passed through a band-pass filter to obtain a logarithmic change in amplitude. The sound is detected by comparing this with a certain threshold value. Also in this case, it is preferable that the band-pass filter passes a few Hz to 20 Hz, which is a change component of the sound.

Since the voice may continue for several seconds after detecting the change in the voice, a timer is provided so that the control signal is continuously output for a predetermined time. The setting time of the timer is desirably about one second to several seconds, since a sufficient time that the steady state of the voice may continue is required.

Further, with respect to the problem that if the AD conversion is performed after the detection, it is not possible to keep up with the rise of the voice. When the power consumption of the AD conversion circuit or the like is allowed, the AD conversion circuit is always operated. Alternatively, a method of storing data for a certain period of time in the past in a buffer memory can be adopted. However, when it is necessary to reduce power consumption, voice detection is performed separately at a threshold lower than the voice detection threshold, AD conversion is started at a timing slightly earlier than voice detection, and data is stored in the buffer memory. It is desirable to store it. Specifically, a power supply controller is provided in the AD conversion circuit in addition to the AD converter, the buffer memory, the interface, and the clock. Then, before the sound processing device connected to the output of the sound detection control device starts operating according to the sound detection control signal, the past data stored in the buffer memory is read to prevent data loss. It is possible.

In the above description, since the entire frequency band of the input signal is used as one amplitude signal, it may not be possible to detect a relatively small sound when there is a loud noise. obtain. On the other hand, in reality, the background noise is rarely white noise existing at the same level in all bands, and is often concentrated in a certain frequency band. As a typical example, the noise in a vehicle during traveling has a very large noise energy component in a low frequency range of several tens Hz or less. For this reason, even when background noise having a larger amplitude than voice exists when observed in all bands, it is often possible to sufficiently observe the voice amplitude in the frequency band excluding the low frequency band. Utilizing this, a filter bank provided with a band-pass filter for each frequency band is provided at the preceding stage, and voice detection is performed on the output for each frequency band as described above, and the presence or absence of voice for each frequency band. Judge. By integrating these results, sufficient detection accuracy can be obtained even in an environment with a large amount of noise by using information in a frequency band where the influence of the noise is small. Instead of providing a band-pass filter and a rectifier, the input signal may be amplitude-modulated at a plurality of different frequencies and then processed by a low-pass filter having predetermined characteristics, thereby converting the input signal into an amplitude signal for each frequency band. It is possible.

Further, the amplitude signal of each band after passing through the filter bank is passed through a low-pass filter and input to an AD conversion circuit via a multiplexer, so that the amplitude signal of each frequency band is converted into digital data in a buffer memory. And send it to the audio processor.

When stationary noise is present, the effect of the noise can be reduced by subtracting the output of the low-pass filter having a lower cutoff frequency from the output of the low-pass filter. is there.

Such a voice detection process can be performed by software.

Hereinafter, the present embodiment will be described with reference to the drawings.

(Embodiment 1) FIG. 1 is a diagram for explaining a configuration of a voice detection control device of Embodiment 1. In the case of audio, an input signal is generally obtained by amplifying an output from a microphone with an amplifier, but may be a signal from an audio device, a television set, a telephone, a wireless device, or the like.

The rectifier 101 comprises a full-wave rectifier circuit or a half-wave rectifier circuit, and rectifies an input signal having a positive or negative value into an amplitude signal having only a positive value. The rectifier can use a circuit that outputs the square of the input.

The low-pass filter 102 has a cutoff frequency of about several Hz to about 1 Hz or less, and extracts only a low-frequency component to obtain a value corresponding to an average of several hundred milliseconds to several seconds of an amplitude signal of an input signal. (Average amplitude value).

The amplifier 103 includes a low-pass filter 102
The average amplitude value obtained in step (1) is amplified and set as a threshold for voice detection. This threshold value affects how much the sound amplitude is detected with respect to noise. In the present embodiment, the amplification factor (gain) can be externally adjusted by using a variable amplifier. However, a fixed amplifier and a voltage dividing resistor which are several times larger may be used. In addition,
It is preferable to adjust the amplification factor of the amplifier 103 in accordance with the bandwidth of the low-pass filter 102 and the band-pass filter 104. However, experiments have shown that it is sufficient to adjust the gain around about 1 times. In the case of 1.times., The amplifier can be omitted.

The band pass filter 104 has a frequency of several Hz to 20 Hz.
It is a filter that passes signals of about Hz. The band-pass filter 104 extracts a component that fluctuates due to voice in the amplitude signal of the input signal.

The rectifier 105 is for detecting whether the fluctuation component of the sound is positive or negative as a positive signal, and can be omitted. The rectifier 10
If the step 5 is omitted, of the fluctuation components of the voice extracted by the band-pass filter 104, the fluctuation in the positive direction, that is, the increase in the amplitude, can be detected as a positive signal. On the other hand, since the signal in the negative direction, that is, the decrease in the amplitude, becomes a negative signal, it cannot be detected by the comparator described later, but the utterance time in which the time for outputting the control signal from the timer described later is estimated. By setting the length sufficiently long, voice can be detected.

The comparator 106 includes the low-pass filter 102
The threshold value obtained by amplifying the output from the amplifier 103 by the amplifier 103 and the output from the band-pass filter 104 are
5, and a comparison with an amplitude change signal obtained when the rectifier 105 is omitted (the amplitude change signal from the band-pass filter 104 when the rectifier 105 is omitted) is used. If the input signal is larger than a certain offset value, it is determined that the input signal may include a voice signal, and a voice detection signal is output. This offset value is set to zero or a small value near zero.

When the timer 107 receives the sound detection signal from the comparator 106, the timer 107 outputs a control signal for a time period in which the sound is estimated to continue. It is considered that about 1 second to about 10 seconds is appropriate as the setting time of the timer 107. With this control signal, the operation and stop of a speech processing device such as a speech recognition device can be controlled.

FIG. 2 is a diagram showing an internal processing waveform in the voice detection control device of this embodiment shown in FIG. 1a to 1e in the drawing show the case where the input signal includes a voice, and 2a to 2e show the case where no voice is included under the same environmental noise. The vertical axis indicates amplitude, and the horizontal axis indicates time.

1a and 2a show input waveforms. The S / N ratio of the voice of the input waveform 1a is about 15 dB.

Reference numerals 1b and 2b denote outputs of the amplifier 103, which indicate threshold values for voice detection. In this example, since the gain of the amplifier 103 is set to 1.0, the same amplitude as the output of the low-pass filter 102 is shown. The low-pass filter 102 was a second-order Butterworth filter with a cutoff frequency of 1 Hz.

1c and 2c are band-pass filters 104
The output of The band-pass filter 104 was a second-order Butterworth filter with a low cutoff frequency of 1 Hz and a second order with a high cutoff frequency of 5 Hz.

1d and 2d are the outputs of the rectifier 105, and are considered to indicate the magnitude of the change in the amplitude of the input signal.

1e and 2e denote the threshold values (1b, 2d) from the amplitude variation (1d, 2d) input to the comparator 106.
The value obtained by subtracting b) is shown. Here, since the offset value of the threshold value is zero, when this value becomes positive, the comparator 106 outputs a voice detection signal indicating that voice has been detected, and outputs a control signal that is ON for a predetermined time from the timer 107. . The difference signal 1e is proportional to the level of the input signal, but it is in principle clear that the sign of the difference signal is exactly the same if the S / N ratio of the input voice is the same.
Therefore, the comparator 106 outputs the same result regardless of the level of the input signal if the S / N ratio of the voice is the same. Therefore, it can be seen that the voice detection control device of the present embodiment operates regardless of the level of the input signal.

(Embodiment 2) FIG. 3 is a diagram for explaining the configuration of a voice detection control device according to Embodiment 2. This voice detection control device includes a voice detection circuit 216 and an AD conversion circuit 2
17.

The basic configuration of the voice detection circuit 216 is almost the same as the configuration of FIG. 1, except that two amplifiers 203 and 208, two comparators 206 and 209, and two timers 207 and 210 are provided. The rectifiers 201 and 205, the low-pass filter 202, and the band-pass filter 204 have the same functions as the rectifiers 101 and 105, the low-pass filter 102, and the band-pass filter 104 shown in FIG.

A control signal for an external audio processing device is created by the amplifier 203, the comparator 206 and the timer 207, and a control signal for the AD conversion circuit 217 is created by the amplifier 208, the comparator 209 and the timer 210. . By setting the amplification factors of the amplifier 203 and the amplifier 208 to different values, a control signal based on two thresholds can be obtained. In order to prevent the beginning of the sound sent to the sound processing device from being lost, the response is speeded up by slightly increasing the high cutoff frequency of the band-pass filter, and the amplifier 20 for the AD conversion circuit 217 is used.
It is desirable to set the threshold value of 8 slightly lower, operate the AD conversion circuit 217 at a timing slightly earlier than the voice detection, and store data in the buffer memory 214 described later. If only the rise time delay of the controlled speech processing device is a problem, the amplifier 208, the comparator 209, and the timer 210 are omitted, and only one control signal is obtained. It may also serve as a control signal for the AD conversion circuit 217. Also, by finding only one control signal for the external audio processing device and constantly operating the AD conversion circuit 217, power consumption is increased, but loss of audio can be completely prevented.

Power supply controller 2 provided in AD conversion circuit 217
11 supplies power to the clock 212 and other circuits only when the control signal is turned on. The AD converter 213 is turned on (when the control signal is turned on).
, And converts the input signal into digital data and stores it in the buffer memory 214. The interface 215 communicates with an external voice processing device and outputs data stored in the buffer memory 214.

(Embodiment 3) FIG. 4 is a diagram for explaining the configuration of a voice detection control device according to Embodiment 3. This voice detection control device includes a plurality of bandpass filters 301 to 30.
3 and a plurality of voice detection circuits 304 to 306 and an OR circuit 31
3 and a timer 314. Rectifier 30
7 and 311, the low-pass filter 308, the band-pass filter 310, the amplifier 309, and the comparator 312 include the rectifiers 101 and 105 and the low-pass filter 10 shown in FIG.
2. It has the same functions as the band-pass filter 104, the amplifier 103, and the comparator 106.

Each of the band-pass filters 301 to 303 has a different center frequency and constitutes a filter bank. The intervals between the center frequencies may be equal, and the logarithm of the center frequency may be equal. Further, a filter bank having a center frequency and a bandwidth in consideration of auditory characteristics such as a mel characteristic and a bark characteristic may be formed. The number of band divisions is not particularly limited. To increase the detection accuracy, it is preferable to increase the number of band divisions. However, from the viewpoint of the circuit scale, it is practical to set the number of bands to about 4 to 16 bands.

For the input signal passed through each band-pass filter, as in the voice detection control device shown in FIG.
Voice detection is performed for each band using the rectifier 307 to the comparator 312.

The OR circuit 312 integrates the sound detection results of the respective bands and, for example, when any one of the bands is turned on (when a sound detection signal is output from the comparator, when the band is turned on, the sound detection signal is turned on). The OR circuit 312 outputs
Does not output ON when one band is turned on, but turns on when multiple bands are turned on at the same time.
May be output.

When ON is input as a sound detection signal from the OR circuit 312, the timer 313 outputs ON as a control signal for a predetermined time.

(Embodiment 4) FIG. 5 is a diagram for explaining the configuration of a voice detection control device according to Embodiment 4. This voice detection control device includes a plurality of bandpass filters 401 to 40.
3 and a plurality of voice detection circuits 404 to 406, an AD conversion circuit 415, an OR circuit 413, and a timer 414. Rectifiers 407, 411, low-pass filter 40
8, the band-pass filter 410, the amplifier 409, and the comparator 412 have the same functions as the rectifiers 101 and 105, the low-pass filter 102, the band-pass filter 104, the amplifier 103, and the comparator 106 shown in FIG. Further, the band-pass filters 401 to 403 and the OR circuit 413
The timer 414 includes the band-pass filters 301 to 303, the OR circuit 313, and the timer 31 illustrated in FIG.
4 has the same function as that of FIG. Further, the clock 417 and the power controller 416 have the same functions as the clock 212 and the power controller 211 in FIG.

The AD conversion circuit 415 converts the amplitude signal of each band output from each sound detection circuit into a multiplexer 42.
The data is converted into digital data by the AD converter 418 while being switched by 2 and stored in the buffer memory 419. The stored data is transferred to an external voice processing device through the interface 420. At this time, a low clock is sufficient for the AD conversion speed since the AD conversion of the entire band data may be performed at intervals of 5 ms to 10 ms.

For AD conversion, aliasing due to sampling (data is distorted by inputting a frequency higher than 1/2 of the sampling frequency,
A low-pass filter is required in order to avoid the phenomenon that accuracy is deteriorated). For this reason, in the present embodiment, in addition to the sound detection and detection circuits shown in FIG.
By providing 1, the high frequency component of the amplitude signal passing through the rectifier 407 is cut off. According to the sampling theorem, the cutoff frequency needs to be half or less of the sampling frequency. Furthermore, in order to eliminate the influence of the pitch cycle of the voice, it is desirable to cut off at a frequency lower than about 50 Hz, which is said to correspond to a male low voice. From the above, in the case of 5 ms sampling, about 50 Hz, 10
In the case of ms sampling, a cutoff frequency of about 25 Hz to about 50 Hz is considered to be sufficient.

In FIG. 5, one control signal is obtained for each band. However, as shown in FIG. 3, an amplifier and a comparator are provided for each band, and two OR circuits and two timers are provided. With this arrangement, two control signals may be obtained and one of the control signals may be used for controlling the AD conversion circuit 415.

(Embodiment 5) FIG. 6 is a diagram for explaining the configuration of a voice detection control device according to Embodiment 5. In the present embodiment, in the voice detection control device shown in FIG.
The adder 432 is connected to the low-pass filter 408 as shown in FIG.
421, the output of the low-pass filter 421, the low-pass filter 4 having a lower cut-off frequency.
08 is subtracted using the adder 432 to obtain an amplitude value with the steady amplitude value of the band set to 0.

This is possible even in the case where stationary noise is present, and as a result, an effect almost equivalent to that obtained by subtracting noise can be obtained. To stabilize the operation, it is also effective to slightly attenuate the output of the constant-pass filter 408 so that the output is not excessively subtracted, or to add a circuit so that the result of the subtraction does not become negative.

(Embodiment 6) FIG. 7 is a diagram for explaining the parts constituting the voice detection control device of Embodiment 6.
The configuration example of a portion that divides an input signal into each band and converts it into an amplitude signal is shown for only one band.

The band amplitude detection circuit 501 shown in FIG.
Comprises a bandpass filter 502 and a rectifier 503.
The band-pass filter 502 and the rectifier 503 correspond to the band-pass filters 301 to 30 shown in FIGS.
3, 401 to 403 and rectifiers 311, 411.

The band amplitude detection circuit 504 shown in FIG.
Is an implementation of the circuit 501 in another configuration. The sine wave oscillator 505 generates a sine wave having the same frequency and a phase difference of 90 degrees. Multipliers 506 and 508 respectively multiply the input signal by two types of sine waves having different phases from sine wave oscillator 505. The multipliers 507 and 509 are connected to the multiplier 5
The result of multiplication of 06 and 508 is squared. Adder 510
Adds two multiplication results. Low-pass filter 511
Extracts only low-frequency components included in the addition result. With the above processing, the oscillation frequency of the sine wave oscillator 505 is set as the center frequency, and the cutoff frequency of the low-pass filter 511 is set to 2
The short-term energy of a signal that has passed through a band-pass filter whose bandwidth is twice as long can be obtained. This output is passed through a band-pass filter (308 in FIG. 4) and a low-pass filter (310 in FIG. 4) similar to those shown in FIG. 4, and compared, thereby performing voice detection in each band.

In the band amplitude detection circuit 501 shown in FIG. 7A, not the energy of each band but the amplitude is obtained. To find the energy for each band, see FIG.
As in the band amplitude detection circuit 512 shown in (c), the amplitude signal for each band converted by the band-pass filter 513 and the rectifier 514 can be squared by the multiplier 515. For voice detection, either the amplitude or the energy of the signal can be used as well.

(Embodiment 7) FIG. 8 is a diagram for explaining the configuration of a voice detection control device according to Embodiment 7. In this embodiment, after the input signal is rectified by the rectifier 601 to be an amplitude signal, the input signal is logarithmized by the logarithmic amplifier 602,
The band-pass filter 603 extracts only the component that varies with time, and the rectifier 604 converts the increase and decrease into positive values. This value is compared with a fixed threshold value by the comparator 605 to detect sound, and the timer 606 generates a control signal.

In the first to sixth embodiments described above,
In order to perform voice detection according to the input range, the threshold is calculated from the average amplitude of the input signal by a low-pass filter. In the present embodiment, however, by using the logarithmic amplifier 602, a fixed threshold is used. It is possible to support various input ranges. Also in the present embodiment, the rectifier 604 can be omitted. Further, logarithmic amplifier 602
Can be changed, but it is also possible to fix it at the stage of designing the device.

(Eighth Embodiment) FIG. 9 is a diagram for explaining the configuration of a voice detection control device according to an eighth embodiment. In the present embodiment, the voice detection processing of the above-described first to seventh embodiments is performed by digital processing.

After the input signal is converted into a digital signal by the AD converter 801, rectification, low-pass filter processing, band-pass filter processing, addition and subtraction by a comparator, and multiplication by an amplifier are performed using a processor 802. And outputs a control signal. As the processor 802, a digital signal processor (DSP), a general-purpose microcomputer (CPU), or a logic circuit can be used. The sampling frequency of audio data is 10
Since the frequency is about kHz to 20 kHz and the processing is relatively simple, it is possible to reduce power consumption by using a low-speed processor.

Further, in order to effectively use the result of the A / D conversion, a RAM 803 is connected, data until a voice is detected is stored in a ring buffer format, and communication with a subsequent voice processing device is performed as necessary. Thus, data loss can be prevented. The buffer capacity for storing audio may be 1 second or less, and the internal memory of the processor may be used.

[0084]

As described in detail above, according to the present invention,
Using analog processing, a relatively simple logic circuit, or a low-speed processor, it is possible to perform sound detection adaptive to the environment and to control an external device by the sound detection. Therefore, by controlling the operation / stop of a relatively high power consumption voice processing device represented by a voice recognition device, a security device or a voice remote control that constantly monitors and processes voice while using low power consumption. Can be realized.

The present invention has a wide range of adaptation to environmental noise, and does not require adjustment by a person from a very quiet environment to a noisy environment. Also, after the voice is spoken,
Since the audio data can be stored in the buffer memory even during the time from when the audio is detected until the processing is started, the data is less likely to be lost, and malfunction due to the loss can be prevented. Furthermore, by performing AD conversion on the output from the filter bank, processing in an external audio processing device can be reduced, and power consumption can be reduced.

[Brief description of the drawings]

FIG. 1 is a diagram for describing a configuration of a voice detection control device according to a first embodiment.

FIG. 2 is a diagram illustrating an example of an internal processing waveform in the voice detection control device according to the first embodiment.

FIG. 3 is a diagram illustrating a configuration of a voice detection control device according to a second embodiment.

FIG. 4 is a diagram illustrating a configuration of a voice detection control device according to a third embodiment.

FIG. 5 is a diagram illustrating a configuration of a voice detection control device according to a fourth embodiment.

FIG. 6 is a diagram illustrating a configuration of a voice detection control device according to a fifth embodiment.

FIGS. 7A to 7C are diagrams illustrating a configuration example of a part of an audio detection control device according to a sixth embodiment, which divides an input signal into respective bands and converts the divided signals into amplitude signals.

FIG. 8 is a diagram illustrating a configuration of a voice detection control device according to a seventh embodiment.

FIG. 9 is a diagram illustrating a configuration of a voice detection control device according to an eighth embodiment.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 101 Rectifier 102 Low-pass filter 103 Amplifier 104 Band-pass filter 105 Rectifier 106 Comparator 107 Timer 201 Rectifier 202 Low-pass filter 203 Amplifier 204 Band-pass filter 205 Rectifier 206 Comparator 207 Timer 208 Amplifier 209 Comparator 210 Timer 211 Power supply control 212 clock 213 AD converter 214 buffer memory 215 interface 216 audio detection circuit 217 AD conversion circuit 301 band-pass filter 302 band-pass filter 303 band-pass filter 304 audio detection circuit 305 audio detection circuit 306 audio detection circuit 307 rectifier 308 low-pass Filter 309 Amplifier 310 Band-pass filter 311 Rectifier 312 Comparator 313 OR circuit 314 Timer 4 01 band pass filter 402 band pass filter 403 band pass filter 404 voice detection circuit 405 voice detection furnace 406 voice detection circuit 407 rectifier 408 low pass filter 409 amplifier 410 band pass filter 411 rectifier 412 comparator 413 OR circuit 414 timer 415 AD Conversion circuit 416 power controller 417 clock 418 AD converter 419 buffer memory 420 interface 421 low-pass filter 422 multiplexer 423 adder 501 band-amplitude detection circuit 502 band-pass filter 503 rectifier 504 band-amplitude detection circuit 505 sine-wave oscillator 506 multiplier 507 Multiplier 508 Multiplier 509 Multiplier 510 Adder 511 Low-pass filter 512 Band-amplitude detection circuit 513 Band-pass filter 514 rectifier 515 multiplier 601 rectifier 602 logarithmic amplifier 603 band pass filter 604 rectifier 605 comparator 606 timer 801 AD converter 802 CPU 803 RAM

Claims (13)

[Claims] 1. A rectifier for rectifying an input signal to convert it into an amplitude signal, and two kinds of different characteristics having different characteristics for individually processing the amplitude signal from the rectifier and obtaining an amplitude change amount and a threshold value. By comparing the output from each of the filter means and the filter means, when the variation in the amplitude exceeds a threshold value, it is determined that there is a possibility that the input signal contains sound, and the sound detection signal is detected. An audio detection control device comprising: a comparing unit that outputs a signal; and a timer unit that outputs a control signal for a predetermined time when an audio detection signal is output from the comparing unit. 2. A rectifier for rectifying an input signal and converting the input signal into an amplitude signal, a logarithmic amplifier for logarithmically converting the input signal converted to an amplitude signal by the rectifier, and processing an output from the logarithmic amplifier. A band-pass filter means for performing the operation, and when the output from the band-pass filter means exceeds a predetermined threshold value, determine that there is a possibility that sound is included in the input signal and output a sound detection signal. An audio detection control device comprising: a comparison unit; and a timer unit that outputs a control signal for a predetermined time when an audio detection signal is output from the comparison unit. A second rectifier for re-rectifying the amplitude signal processed by the band-pass filter, wherein an output from the band-pass filter is input to the comparator via the second rectifier. The voice detection control device according to claim 1 or 2, wherein: 4. An amplifying means for amplifying an amplitude signal processed by the low-pass filter means, wherein an amplification factor of the amplifying means can be adjusted from the outside, and an output from the low-pass filter means is output from the amplifying means. The voice detection control device according to claim 1, wherein the voice detection control device is input to the comparison unit via a control unit. 5. An analog-to-digital converter for converting the input signal into digital data, and a buffer memory, wherein the input signal before voice detection is stored as digital data in the buffer memory. The voice detection control device according to the above. 6. A plurality of amplifying means for amplifying an output from said low-pass filter means, a plurality of comparing means, and a plurality of timer means, and a plurality of control signals are set by setting amplification factors of the respective amplifying means to different values. 6. The voice detection control device according to claim 5, wherein a power supply and a clock of the analog-to-digital converter and the buffer memory are controlled by one of them. 7. A plurality of second band-pass filter means for separating an input signal for each frequency band, and a plurality of second band-pass filter means for converting each band signal separated by each second band-pass filter means into an amplitude signal. And a plurality of sets of different characteristics, each of which separately processes an amplitude signal from the rectifier and obtains an amplitude change amount and a threshold.
The type of filter means and the output from each filter means are compared for each frequency band, and when the amount of change in amplitude exceeds a threshold value, there is a possibility that voice may be included in the input signal. A plurality of comparing means for judging and outputting a sound detection signal; a total means for summing the sound detection signals from the respective comparing means; and the timer for outputting a control signal for a predetermined time based on an output from the synthesizing means. The voice detection control device according to claim 1, further comprising: 8. A plurality of second band-pass filter means for separating an input signal for each frequency band, and a plurality of band signals for converting each band signal separated by each second band-pass filter means into an amplitude signal. A plurality of logarithmic amplifiers that logarithmically convert an amplitude signal for each frequency band from the rectifiers, and a plurality of bandpass filters that process outputs from the logarithmic amplifiers for each frequency band. Means for outputting a voice detection signal by determining that voice may be included in an input signal of the frequency band when an output from the band-pass filter means exceeds a predetermined threshold. Said comparing means, comprehensive means for integrating the voice detection signals from each of the comparing means, and the timer means for outputting a control signal for a predetermined time based on the output from the comprehensive means. The voice detection control device according to claim 2. 9. A second low-pass filter for processing an amplitude signal for each frequency band converted by the second band-pass filter and the rectifier, and 9. The voice detection according to claim 7, further comprising: a multiplexer and an analog-to-digital converter for converting an output of the digital signal into digital data; and a buffer memory, wherein an amplitude signal for each frequency band is stored as digital data in the buffer memory. Control device. 10. An amplitude signal for each frequency band converted by said second band-pass filter means and said rectifying means, averaged by said low-pass filter means, is subtracted from said amplitude signal for each frequency band. The voice detection control device according to claim 9, wherein the digital data is converted into digital data after reducing the influence of noise. 11. A plurality of amplitude modulation means for amplitude-modulating an input signal at a plurality of different frequencies, and a plurality of third low-pass filter means each processing output from each of the amplitude modulation means and having predetermined characteristics. With
Further, the amplitude signal for each frequency band converted by the amplitude modulation means and the third low-pass filter means is individually processed, and a plurality of sets of two different characteristic sets for obtaining an amplitude change amount and a threshold value, respectively. The type of filter means and the output from each filter means are compared for each frequency band, and when the amount of change in amplitude exceeds a threshold value, there is a possibility that voice may be included in the input signal. A plurality of comparing means for judging and outputting a sound detection signal; a total means for summing the sound detection signals from the respective comparing means; and a timer means for outputting a control signal for a predetermined time based on an output from the summing means. A voice detection control device comprising: 12. A plurality of amplitude modulating means for modulating the amplitude of an input signal at a plurality of different frequencies, and a plurality of third low-pass filter means each processing output from each of the amplitude modulating means and having predetermined characteristics. With
A plurality of logarithmic amplifiers for logarithmically converting the amplitude signal of each frequency band from the rectifier; and a logarithmic conversion of the amplitude signal for each frequency band converted by the amplitude modulator and the third low-pass filter. A plurality of logarithmic amplifying means for converting, a plurality of bandpass filter means for processing an output from the logarithmic amplifying means for each frequency band, and when an output from the bandpass filter means exceeds a predetermined threshold, A plurality of comparing means for judging that there is a possibility that sound is included in the input signal of the frequency band and outputting a sound detection signal; a total means for summing the sound detection signals from the respective comparing means; A voice detection control device comprising: timer means for outputting a control signal for a predetermined time according to an output from the synthesis means. 13. The voice detection control device according to claim 1, further comprising an analog-to-digital converter and a digital processor, wherein the input signal is digitally processed by software.