JPH0396999A - Sound collecting device - Google Patents

Abstract

PURPOSE:To collect sound with a small noise component by arranging a main acoustic input means and plural subordinate acoustic input means at respective necessary positions and erasing acoustic signals of the subordinate acoustic input means from an acoustic signal of the main acoustic input means. CONSTITUTION:A microphone 21 as the main acoustic input means is arranged nearby on a microphone array plate 11 nearby a object sound source of sound collection and microphones 22 - 28 as the subordinate acoustic means are arranged at distributed positions around the microphone 21. An electric signal corresponding to the main sound outputted by the microphone 21 is processed by an A/D converter and an adaptive filter and electric signals corresponding to subordinate sounds from the microphones 22 - 28 are processed similarly; and the total of the subordinate sound processed data is subtracted from the main sound processed data to realize small-noise, high-quality sound collection.

Description

Detailed Description of the Invention (Field of Industrial Application) The present invention relates to a device for collecting sound emitted from a sound source at a specific location, and in particular to a device for collecting sound9 (noise) emitted by many other sound sources. The sound generated by the sound source at the specific location (main sound)
This invention relates to a sound collection device that selectively collects sound with high gain.

(Prior art) For example, when a driver operates various in-vehicle devices in a vehicle, and there are many operable switches, he searches for the position of the switch to be operated and positions his finger on the switch. It is necessary to visually check up to
While the vehicle is running, controlling various in-vehicle devices by the driver's manual operation poses a problem in terms of vehicle driving safety.

Therefore, in order to eliminate the need for visual recognition and finger operations, a voice recognition device detects the driver's vocalizations (main sound), identifies vocal commands, and controls in-vehicle equipment in response to the commands. It is proposed that.

However, especially in a car, the driver's voice (tonic Il+) is accompanied by external noise (human input to the voice recognition device).
The level of noise (noise) may be large and may seriously impede speech recognition. Examples of external noise include sounds from outside the vehicle, the sound of the vehicle's engine, sounds from a sound system inside the vehicle, and voices of fellow passengers.

Therefore, when a specific switch is operated, in preparation for voice input, a noise reduction device is installed that lowers the output of the on-board acoustic system and closes the window glass in order to reduce the level of ML sound. - Proposed in Publication No. 29755. In addition, the present inventors have provided a sound collection device that clearly extracts the driver's voice from audio speaker sounds and engine sounds (Japanese Patent Application No. 1-77776,
-79512).

(Problem to be Solved by the Invention) In devices where a specific switch operation is performed before voice input in order to reduce the noise level, this operation is troublesome and it is necessary to wait until the window glass is completely quiet. However, there are problems such as it takes a long time to start voice input.

In addition, the sound collection device proposed in the aforementioned Japanese Patent Application No. 177776 and Japanese Patent Application No. 1-79512, which performs sound W signal processing on noise from speakers, engines, etc., detects sound sources such as audio equipment located at specific locations. It is possible to suppress noise and separate the audio signal (main sound) with high gain for the sound (noise) emitted, so-called fixed noise, but it is difficult to suppress noise from noise from sound sources outside the vehicle such as oncoming cars, that is, moving noise. has a low noise suppression effect, and the audio signal (tonic 1i
P) high gain separation is not sufficient. In other words, the main sound emitted by the target sound source can be collected with high gain in response to noise generated by a noise source at a specific position.
The noise suppression effect is low for noise generated by other noise sources.

An object of the present invention is to collect a main sound emitted by a sound source to be collected with high gain, not only from noise generated by a noise source at a specific position but also from noise generated by 34 other noise sources.

[Structure of the invention]

(Means for Solving the Problems) A sound collection device of the present invention is located near a sound source to be collected,
Main sound input means for converting f-sound into electric signal (.2)
i A plurality of auxiliary sound input means (22 to 28) for converting sound into electrical signals, distributed around the main sound input means (21) and surrounding the main sound input means (21); acoustic signal processing means (
ioo) ; Note that symbols inside brackets indicate corresponding elements in the embodiments shown in the drawings and described later.

(Function) The auxiliary sound input means (22 to 28) are the main sound input means (21
) are distributed around the main sound input means (21), so that sounds (noise) emitted by sound sources other than the sound source to be collected first pass through the auxiliary sound input means (22 to 22).
28) and reaches the main sound input means (2l), so when the main sound input means (2l) generates an electric signal corresponding to noise, the sub-sound input means Pi (22 to 28) generate an electric signal corresponding to the generated noise. The electric signal is processed by the signal processing means (10
0) has been reached. For this reason, the acoustic signal processing means (i
oo) outputs an acoustic signal in which the electric signal generated by the sub-acoustic input means (22 to 28) is substantially eliminated from the electric signal generated by the main acoustic input means (2l), so the output acoustic signal is , the noise component is reduced.

On the other hand, the main sound emitted by the sound source to be collected first reaches the main sound input means (2I) and then, after a certain time delay, reaches the sub-sound input means (22-28). The electrical signals generated by ~28) have a certain delay with respect to the electrical signals generated by the main sound input means (21), so that the electrical signals corresponding to the main sound generated by the main sound input means (2I) is output as an acoustic signal, although the waveform is disturbed to some extent.

As a result, noise, whether it is emitted from a fixed sound source or a moving sound source, is greatly suppressed in the output acoustic signal, and the sound collection gain of the main sound generated by the sound source to be collected is reduced. High L).

In order to further increase this sound collection gain, in a preferred embodiment of the present invention, the main sound input means (2I) has directivity toward the sound source of the input object; the auxiliary sound input means (22-28)
has directivity in a direction opposite to the direction toward the main sound input means (2l); The difference between the electricity generated by means (22 to 28) is the real difference.
Adaptive filter (62~
68), and a subtraction means (70) for outputting an acoustic signal obtained by subtracting the electric signal generated by the adaptive filter (62-68) from the electric signal generated by the main acoustic input means (21).

According to this, the main sound input means (21) converts the main sound generated by the sound source to be collected into a high-level electrical signal, and the auxiliary sound input means (22 to 28) convert it into a low-level electrical signal. Therefore, by subtracting the electrical signal of the main sound generated by the auxiliary sound input means (22 to 28) from the electrical signal of the main sound generated by the main sound input means (21), the main sound in the output sound signal can be separated. level drop and waveform disturbance are reduced. Conversely, the level of noise decreases. Furthermore, since the adaptive filter includes a delay means, its output is delayed with respect to the input. Corresponding to the delay time of this delay means,
When noise arrives at the sub-sound input means (22 to 28) and reaches the main sound input means (21), the adaptive filter outputs an output corresponding to the noise.
! ! ! II1 acoustic input means (21) for input means (21)
By setting the distances 22 to 28), the noise component in the output acoustic signal becomes extremely low, and the sound collection gain of the main sound becomes significantly high.

Other objects and features of the invention will become apparent from the following description of embodiments with reference to the drawings.

(Embodiment) FIG. 1 shows an outline of the interior appearance of an automobile equipped with an embodiment of the present invention. In this embodiment, a system is provided in which the navigation device is operated by a driver's voice input.

Referring to FIG. 1, the car body IO has a navigation device monitor television 13 built into the center console, and a navigation unit 1-l2, which is the navigation device body, is installed in a hidden position near it. It is being Further, a microphone array board 11 is attached to a portion of the ceiling 14 above the driver's seat. Although not shown in the figure, multiple audio speakers are installed inside the vehicle.

FIG. 2 shows a view of the microphone array plate II from above the vehicle.

A main input microphone 21 is attached to the microphone array board 11 at a position slightly closer to the driver's seat side of the windshield glass than the center of the board, that is, a position directly above the mouth of the seated driver. A plurality of sub-input microphones 22 to 28 are attached at a predetermined distance on the same circumference surrounding the input microphone 21 at a predetermined distance. Note that originally, another sub-input microphone was placed between the sub-input microphones 22 and 28, and the front and rear,
It is preferable to arrange the sub-input microphones symmetrically in left-right and diagonal directions, but since the sub-input microphones 22 and 28 are located near the right end of the vehicle interior, and there is no space to arrange the other sub-input microphone, This installation is omitted. 29 in the diagram
indicates the microphone cords of microphones 21 to 28.

Figures 3a, 3b, and 3C show the positional relationship between the driver's head and the microphone when the driver is sitting in the driver's seat, and Figure 3a is a view from the front of the car.
Figure 3b is a view from the side of the car, and Figure 3c is a view from above the car. Note that the arrow shown in FIG. 3C indicates the directivity direction of the sub-input microphone. Main input microphone 21
has directivity downward in the vertical direction, whereas sub-input microphones 22 to 28 have directivity in a direction opposite to the direction toward main input microphone 21.

The microphone array board 11 is installed in the car 1 so that the main input microphone 21 is located slightly in front of the driver's head and directly above the driver's head.
It is attached to the ceiling 14 of 0.

In other words, one driver (sound rA) and two main input microphones]
The distance between the driver's mouth (sound source) and the sub-input microphone 21 is
It is shorter than the distance of ~28.

In this way, the main input microphone 21 and the sub input microphones 22 to 28
, the driver's voice reaches the main input microphone 2 earlier than it reaches any of the sub-input microphones 22 to 28.

FIG. 4 shows a main input microphone 21 and a sub-input microphone 22 to
28 shows the configuration of the noise canceling unit 100 connected to 28. This noise canceling Unisol l-10
0 indicates that the main input microphone 21 receives an input earlier than any of the sub input microphones 22 to 28.
The signal from the 5th stage is transmitted to the voice recognition unit 200 connected to the 5th stage.

The configuration and operation of the noise cancellation unit 100 will be described below with reference to FIG. 4.

The respective acoustic signals outputted by the input microphones 21 to 28 are passed through band pass filters 3 to 38 and AMP41 to
The signal is inputted to the A/D converter 50 via A/D converter 48, where it is A/D converted. Of these converted signals, only the signals from the sub-input microphones 22 to 28 are filtered through the adaptive filters 62 to 28.
Sent to 68.

Each of the a corresponding filters 62 to 68 includes a 127-stage delay element (Z1-2 delay element) and a 128-stage variable amplification element (
A+~A128: multiplier) and 127 stages of addition elements (
The characteristics of this filter can be varied in various ways by adjusting the coefficients set for each of the variable amplification elements A1 to A128.

The coefficient control circuits 62a to 68a output the output of the subtracter 70, that is, the output dk obtained by A/D converting the signal from the main input microphone 2l, and the signal yk2 output by the adaptive filters 62 to 68.
~''/k8 is input, a group of coefficients whose root mean square value is the minimum is generated, and these coefficients are applied to each of the variable amplification elements A~11 12 AI28. That is, the subtracter 70 The gain of each amplification element is controlled so that the output at is minimized.In this example, the coefficient control circuits 62a to 688 are configured to perform a known algorithm.

Incidentally, the signals output from each input microphone have different times and amplitudes. This difference can be eliminated by passing the signals input to the sub-input microphones 22-28 through adaptive filters 62-68. However, since there is no adaptive filter or delay circuit on the output line of the main input microphone 21, the sound that is input earlier than the input to any of the sub-input microphones 22 to 28 is directly transmitted to the next stage. Recognition unit 200 is reached.

The output components of the main input microphone 21 corresponding to the output components of the sub-input microphones 22 to 28 are obtained by subtracting the outputs of the adaptive filters 62 to 68 from the output of the main input microphone 21 by a subtracter 70, so that the output components of the sub input microphones 22 to 28 can be obtained by subtracting the output components of the main input microphone 21 from the output components of the main input microphone 21. 200 is not really given. In other words, noise from the outside (outside the circular area formed by the sub-input microphones 22-28 arrangement), regardless of whether it is fixed noise or moving noise, is input to any of the sub-input microphones 22-28. The signal is input to the main input microphone 21. noise cancellation unit 100
The outputs of the auxiliary input microphones 22 to 28 are delayed by adaptive filters 62 to 68 to match the timing of the output of the main input microphone 2l.
The circuit operates so that the signals input at 28 are canceled. Therefore, the sound (noise) input to the sub input microphones 22 to 28 before the main input microphone 21 is not substantially output to the speech recognition unit 200.

Second, as described above, only the driver's audio signal is substantially extracted by the arrangement of the main and sub microphones and the noise canceling unit 100 and output to the audio recognition unit 200. Note that this voice recognition unit 200 executes a known algorithm to recognize the voice uttered by the driver, and further uses the recognized voice information and a predetermined navigation command word (for example, regarding the display of a road map, etc.). (enlargement, reduction, on, off, top, bottom, right, left, etc.), and if they match,
A command signal corresponding to the command word is output to the navigation unit 12.

Therefore, in this example, the driver's voice input can give instructions to the navigation unit 1-12 and control it.

Note that in this embodiment, the coefficient control of the adaptive filter is performed using the LMS algorithm; however,
Other algorithms, such as learning identification methods, may also be used.

In addition, in this embodiment, 141 input microphones are arranged on a flat surface, but if they are located further from the audio source than the main microphone and are arranged in a shape surrounding the main input microphone, they can be arranged in three dimensions. But that's fine.

Furthermore, although the above embodiment is a mode in which the voice of the driver is collected by the vehicle, the object of sound collection is not limited to voice, but also the operation of a machine to be monitored for monitoring the operating state of the machine. It may be a sound.

〔Effect of the invention〕

As described above, according to the present invention, moving noise, including fixed noise coming from a certain direction, is sufficiently reduced, and only the sound emitted by the sound source to be collected is clearly extracted.

[Brief explanation of drawings]

FIG. 1 is a perspective view showing the exterior of the interior of an automobile in which an embodiment of the present invention is installed. FIG. 2 is a top view showing the arrangement of microphones on the microphone array plate 1 shown in FIG. 1. FIG. 3a shows the microphone array plate 11 shown in FIG. t.
FIG. 3b is a side view, and FIG. 3c is a top view showing the orientation direction of the microphone on the array plate 1J. FIG. 4 is a block diagram showing the configuration of the noise canceling unit 100 connected to the microphones of the microphone array board 11 shown in FIG. 10: Vehicle body 1l: Microphone array board 15 16 12: Navigation unit l-13 monitor TV 14: Ceiling 21: Main input microphone (main sound input means) 22-28: Sub input microphone (sub sound input means) 29: Microphone coat 3l-38: Bandpass filter 41-48:
Amplifier 50: A/D converter 62-68: Adaptive filter 628-68a
:Coefficient control circuit 70:Subtractor 100:Noise cancel unit 1- (signal processing means)
200: Voice recognition Uninote

Claims (2)

[Claims] (1) Main sound input means located near the sound source to be collected and converts sound into electrical signals; a plurality of auxiliary sound input means for converting into signals; and an audio signal processing means for outputting an audio signal in which the electric signal generated by the auxiliary sound input means is substantially eliminated from the electric signal generated by the main sound input means; Sound collection device. (2) The main sound input means has directivity toward the sound source of the input target; the auxiliary sound input means has directivity in the opposite direction to the direction toward the main sound input means; , an adaptive filter that generates an electric signal that substantially minimizes the difference between the electric signal generated by the main sound input means and the electric signal generated by the auxiliary sound input means, and a filter that is more adaptive than the electric signal generated by the main sound input means. 2. The sound collecting device according to claim 1, further comprising: subtracting means for outputting an acoustic signal obtained by subtracting an electrical signal generated by the sound collector.