JPH0314009A - Sound source detector - Google Patents

Abstract

PURPOSE:To always direct sound detectors toward a sound source by using the fuzzy control to compare detection values obtained from a pair of rotatable sound detectors. CONSTITUTION:A pair of sound detectors 2 to 5, rotating mechanism parts 6 and 7, an angle detector 8, and a fuzzy control part are provided, and this control part turns sound detectors 2 to 5 to such positions by fuzzy inference that the difference of sound volume and/or arrival time between both detected sound values is zero. Thus, sound detectors are always directed toward the sound source, and the direction of the sound source is sufficiently detected even in case of a single spoken voice, a sound changing with time, or a low-frequency sound of low directivity.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of the Invention) The present invention relates to a sound source detection device that can be used in automatic tracking devices for sound collection microphones, various customer service devices, and the like.

(Summary of the Invention) The present invention uses fuzzy control to compare detected values obtained from a pair of rotatable sound detectors so that the sound detectors always face the direction of the sound source.

(Prior art and its problems) Conventionally, as a method of detecting the direction of a sound source, it is known that the direction of the sound source can be detected by rotating a directional microphone equipped with a horn and detecting the peak of the output level. It is being

However, with this method, it is impossible to detect sounds that are uttered singly or sounds that change over time, and it is also difficult to detect low-frequency sounds with low directivity.

In addition, the device itself that implements this method mechanically rotates a microphone equipped with a horn, so its appearance is not desirable, and it is hoped that a device with a simpler and more stable appearance will emerge. It was rare.

(Object of the Invention) This invention was made to solve the above problems, and its purpose is to reduce the sound that is uttered singly,
To provide a sound source detection device capable of sufficiently detecting the direction of the sound source even for temporally changing sounds or low-frequency sounds with low directivity, and capable of simply forming an external shape. There is a particular thing.

(Structure and Effects of the Invention) In order to achieve the above object, the present invention includes a pair of audio detectors arranged approximately symmetrically with respect to a rotation axis, and a pair of audio detectors connected to the rotation axis to rotate the audio detector. A rotating mechanism section to be driven, an angle detector for detecting the rotational position of the rotating shaft, a processing section that compares detected values of the audio detector and calculates a difference in volume, arrival time, etc. between the two, and the angle detector. From the output of the processing section and the output of the processing section, a rotation mechanism section for rotating the sound detector to a position where the volume difference and/or arrival time difference between both detected sound values becomes zero is determined by fuzzy inference using a membership function. The present invention is characterized in that it includes a fuzzy control section that calculates the drive amount.

In this way, the present invention uses fuzzy control to compare detected values obtained from a pair of rotatable audio detectors,
Since the sound detector is always directed toward the sound source, it is possible to detect the direction of the sound source sufficiently even for single vocalizations, temporally changing sounds, or low-frequency sounds with low directionality. Moreover, the external appearance can be configured into a simple and stable shape.

Furthermore, by using fuzzy control in the control section, the configuration of the device is simplified, and the processing speed is also increased, resulting in better responsiveness to changes in the sound source.

(Explanation of Examples) Next, embodiments of the present invention will be described based on the drawings.

FIG. 1 is a plan view of a sound source detection device according to the present invention, and FIG. 2 is a front view of the same.

As shown in the figure, this sound source detection device has a horn 2 for collecting sound at a position corresponding to the ear of a rotating body 1 imitating a human head.
．． 4, and a microphone 3.5 is built into the rotating body 1 at the base thereof. The horn 2.4 and the microphone 3.5 are arranged symmetrically, and as shown in FIG. 1, the upper part of the rotating body 1 is the front side when detecting the sound source.

Further, the rotating body 1 is rotatably supported on a pedestal 9 by a support shaft 7, and the support shaft 7 is connected to a motor 6 installed below the pedestal 9 to be rotationally driven. Furthermore, an angle detector 8 is fitted in the middle of the support shaft 7 to detect the rotational position of the rotating body 1.

As shown in Fig. 1, by placing the horn 2.4 and microphone 3.5 on the rotating body l, if the sound source S moves from the front position to the rear, it will be detected by the microphone 3.5. The difference in volume between the two changes as shown in Figure 4.

That is, at the front, the volumes detected by the microphones 3.5 are the same, and the volume difference (2) is zero. As the sound source S rotates clockwise, the detected volume of the microphone 3 becomes relatively larger than the volume of the microphone 5, and a volume difference occurs at a position 90 degrees directly to the side.
is the maximum.

As it rotates further, the volume difference ■ decreases and becomes zero at the rear.

Similar characteristics can be obtained when the sound source S is rotated counterclockwise. In that case, contrary to the above case, the volume of the microphone 5 will be higher.

Similarly, when the sound BS moves around the rotating body 1, the time difference ψ between the sound emitted from the sound source S and reaching the microphone 3.5 changes as shown in FIG.

That is, at the front, the time taken for the sound to reach the microphone 3.5 is equal, and the arrival time difference ψ is zero.

As the sound 'aS rotates clockwise, the time it takes to reach the microphone 5 becomes relatively slower than the time it takes to reach the microphone 3, and the difference ψ becomes maximum at a position slightly before 90 degrees directly horizontally.

As it rotates further, the difference φ decreases and becomes zero directly behind it.

Similar characteristics can be obtained when the sound source S is rotated counterclockwise. In that case, the time it takes for the signal to reach the microphone 5 will be faster.

It goes without saying that similar characteristics can be obtained even when the rotating body 1 is rotated while the sound source S remains fixed.

FIG. 3 is a block diagram showing a control section of the sound source detection device.

The position angle θ of the rotating body 1 detected by the angle detector 8 is sent to the differentiator 11 and the sound source direction inference section 13.

The differentiator 11 detects the amount of change in the input position angle θ, that is, the rotational speed Δθ, and sends it to the sound source direction inference section 13.

The conversion unit 12 compares the output levels of the microphones 3.5, calculates the volume difference (2) between the two microphones, the arrival time difference ψ, the frequency r, and sends the volume values to the sound source direction inference unit 13.

The sound source direction inference unit 13 uses fuzzy inference from the various input signals to calculate the driving angle Mθ of the motor 6 to direct the rotating body l in the direction of the sound aS, and the direction angle indicating the position of the sound source S. and output.

The output drive angle Mθ is sent to the motor 6 to rotate the rotating body 1 by a predetermined angle.

By performing these series of processes continuously, the front of the rotating body 1 always faces the direction of the sound source S, and the detection level of the microphone 3.5 is maximized. Note that the sound detected by the microphone 3.5 at this time can be directly collected as high-quality stereo sound.

In addition to displaying the direction angle of the sound source S output from the sound source direction inference unit 13 on a sound source angle display as numerical data, it is also possible to rotate other main microphones, close-up cameras, etc. in the direction of the sound source. It is also used to cause

FIG. 6 shows the process of determining the drive angle Mθ of the motor 6 in the sound source direction inference unit 13 shown in FIG. It is a block diagram explaining.

In the figure, the conversion unit 12 compares the output levels of the microphones 3.5, calculates the volume difference V and the arrival time difference φ between the two microphones, and sends them to sample-and-hold circuits 14.15, respectively.

The angle detector 8 sends the detected position angle θ of the rotating body l to the differentiator 11 and the sample-and-hold circuit 16.

The differentiator 11 detects the amount of change in the input position angle θ, that is, the rotational speed Δθ, and outputs the sample and hold circuit 17.
send to

The sample and hold circuits 14 to 17 hold the input volume difference (2) between both microphones, arrival time difference ψ, position angle θ, and rotational speed Δθ at predetermined intervals, and send them to the fuzzy inference device 18 as signals x1 to x4.

The fuzzy inference device 18 calculates the optimal output value based on the fuzzy rule of the block 19 according to the fitness degree (membership value) of the membership function of the input signals x1 to x4, and uses the result. The output signal y1 obtained by defuzzification is sent to the amplifier 20.

The fuzzy rule stored in block 19 uses the values of the input volume difference V(xi), arrival time difference ψ(x2), position angle θ(x3), and rotational speed Δθ(x4) as antecedents, The value of the drive angle Mθ (y 1) of the motor 6 is multiplied by the consequent part, and the input volume difference V (xi), arrival time difference ψ (X2), position angle θ (x3), and rotation speed Δ
Drive angle M of the motor 6 according to the combination of values of θ (X4)
The value of θ(yl) is defined respectively.

The amplifier 20 amplifies the signal y1 and then sends it to the motor 6 as a drive angle Mθ.

FIG. 7 shows the input X1 input to the fuzzy inference device 8.
It is a graph showing a membership function used when fuzzifying ~X4 and further defuzzifying the calculation result to a specific output y1.

Figure a shows a membership function representing the degree of compatibility between the input xi representing the volume difference between the two microphones and each fuzzy label.

The figure shows the input χ2 indicating the arrival time difference ψ and the membership function indicating the degree of compatibility of each fuzzy label.

Figure C is a membership function representing the degree of compatibility between the input x3 indicating the position angle θ and each fuzzy label.

Figure d shows a membership function representing the degree of compatibility between each fuzzy label and the input x4 indicating 0 to the rotation speed.

The horizontal axis of each of these figures represents specific detected values.

Further, Figure e is a membership function regarding the output yt indicating the drive angle Mθ.

These membership functions are optimally set depending on the specific installation interval of the microphones 3.5, detection sensitivity, driving speed of the motor 6, etc. in the device.

In this sound source detection device configured in this way, for example,
The volume difference between both microphones 3 and 5 is exactly large (x 1-P
L) , and the arrival time difference ψ is exactly medium value (x
2=PM) and the rotational speed Δθ of the rotating body 1 is zero (x4=ZR), the drive angle Mθ is 90 degrees in the negative direction (
y 1 =NM) is output, and the motor 6 is driven to rotate 90 degrees counterclockwise.

Similarly, when the volume difference ■, the arrival time difference ψ, and the rotational speed Δθ change, MIN-MAX calculations are performed based on the corresponding multiple fuzzy rules, and the optimal drive angle of the motor 6 is output. be done.

In this way, the position and angle of the sound source S can be estimated from the detected values regarding the sound source S obtained from a pair of microphones, so even for a sound source that emits a single sound or a moving sound source S,
fully detectable.

Also, based on the position angle of the sound source S obtained in this way, the motor 6 is driven, so that the rotating body 1 always faces the direction of the sound source S, and the detection level of the microphone 3.5 for the sound -aS is maintained at its best.

Furthermore, since angular data regarding the direction of the sound source S detected at this time is also output, tracking of the sound source S becomes easier for other devices and equipment.

Further, in this embodiment, since the combination of the rotating body 1 and the horn 2.degree. 4 imitates a human head, a simple and stable external shape can be achieved.

Regarding the sound source direction inference unit 13 shown in FIG.
It is also possible to use a normal microcomputer instead of a fuzzy inference device.

[Brief explanation of drawings]

FIG. 1 is a plan view showing an embodiment of the sound source detection device according to the present invention, FIG. 2 is a front view thereof, FIG. 3 is a block diagram showing the control section of the sound source detection device, and FIG. FIG. 5 is a characteristic diagram showing the relationship between the angle and the detected sound volume difference, FIG. 5 is a characteristic diagram showing the relationship between the angle of the rotating body and the detected arrival time difference, and FIG.
The figure is a block diagram showing details of the sound source direction inference section, and FIG. 7 is a graph showing membership functions set in the fuzzy inference device. 14-17...Sample/hold circuit 18...Fuzzy inference device 20...Amplifier S...Sound source ■...Volume difference ψ...Arrival time difference θ...Position angle Δθ...Rotation Speed Mθ... Drive angle 1... Rotating body 2.4... Horn 3.5... Micro... Motor 7... Support shaft 8... Angle detector 9... Frame 11 ... Differentiator 2 ... Conversion section 13 ... Sound source direction inference section

Claims (1)

[Claims] 1. A pair of sound detectors arranged approximately symmetrically with respect to a rotation axis; a rotation mechanism unit connected to the rotation shaft and rotationally driving the sound detector; An angle detector that detects the rotational position and a processing unit that compares the detection values of the audio detector and calculates the volume difference, arrival time difference, etc. between the two, and a membership function is calculated from the outputs of the angle detector and the processing unit. a fuzzy control unit that calculates the drive amount of the rotation mechanism unit for rotating the audio detector to a position where the difference in volume and/or the arrival time difference between the detected audio values is zero, based on the fuzzy inference used; A sound source detection device characterized by comprising: