JP2000312395A - Microphone system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enhance the S/N of a voice signal in a noise reduction system where two microphones are in use. SOLUTION: In the microphone system that uses output signals of two microphones to output a talker voice signal with the enhanced S/N through adaptive signal processing, the two microphones 11, 12 with directivity are closely arranged and an angle between the directivity of the microphone and a voice utterance direction of a talker differs between the microphones. For example, each of the microphones 11, 12 is mounted on a sunvisor 13 of a vehicle while differentiating the directivity or mounted on a driver's assistant seat or a ceiling at a driver's seat of the vehicle.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a microphone system and, more particularly, to a speaker voice signal having an improved SN ratio by performing adaptive signal processing using signals output from two microphones having directional characteristics with respect to sound receiving sensitivity. The present invention relates to a microphone system that outputs a signal.

[0002]

2. Description of the Related Art Current speech recognition systems have a
If the SN ratio (S: voice / N: noise) is secured,
The technology level has reached 95% recognition rate. However, when the S / N ratio decreases due to noise existing in the surroundings, the recognition rate sharply decreases. FIG. 9 shows the relationship between the SN ratio and the recognition performance for several types of microphones (omnidirectional, unidirectional, narrow directivity, AMNOR (Adaptive Microphone-array for Noise Red).
auction)), and the SN ratio and the recognition rate are generally included in a band showing the S-characteristic 100. As is clear from FIG. 9, the recognition rate sharply decreases due to the decrease in the SN ratio, and decreases to about 50% in an environment where the SN ratio is 0 dB.

[0003] Therefore, the deterioration of the recognition performance described above is unavoidable in a vehicle cabin where noise (engine sound, road noise, pattern noise, wind noise, etc.) generated by the vehicle is present, and the voice recognition system is mounted on the vehicle. This is one of the major problems in making From the circumstances described above, reduce the influence of noise existing around,
Various systems for receiving sound with a high SN ratio have been proposed, and a high SN ratio sound receiving system using a plurality of microphones and digital signal processing is one example. The simplest configuration of such a high S / N ratio sound receiving system is a system as shown in FIG. 10 using two microphones.
Other more advanced systems have been proposed, such as Griffith-Jim arrays and AMNOR.

[0004] In FIG. 10, 1 and 2 first, second microphone, 3 is an adaptive signal processing unit, the output signal x ₂ microphones 2 with the error signal e is input is input as a reference signal, Adaptive signal processing is performed based on the LMS (Least Mean Square) algorithm so that the power of the error signal e is minimized. In the adaptive signal processing unit 3, 3a
Is an LMS operation unit, and 3b is an adaptive filter having, for example, an FIR digital filter configuration. The LMS operation unit 3a determines coefficients of the adaptive filter 3b so that the power of the error signal e is minimized by the adaptive signal processing.

[0005] A target response setting unit 4 receives a signal output from the microphone 1 as a target signal, and satisfies causality. When a signal delay time which is a half of the tap length of the adaptive filter 3b is d, the target response setting unit 4 has a delay characteristic of the time d, and has a flat characteristic (gain 1 characteristic) in an audio frequency band. That is, the target response setting unit 4 has a flat frequency characteristic of a gain 1 as shown in FIG.
Has an impulse response characteristic having a delay time d as shown in FIG. A subtraction unit 5 subtracts the output signal of the adaptive filter 3b from the target response output from the target response setting unit 4 and outputs an error signal e.

During non-speech recognition, only noise is input to the microphones 1 and 2, and the adaptive signal processing unit 3 adjusts the filter coefficient W so that the power of the error signal e, that is, the noise output is minimized by the adaptive signal processing. decide. On the other hand, at the time of speech recognition, the adaptive signal processing unit 3 does not update the filter coefficient, but sets the filter coefficient W determined at the time of non-speech recognition to the adaptive filter 3b and outputs a speech signal. The ideal performance originally required of the system shown in FIG. 10 is to output only the audio signal Xs (z) (noise output is 0) as an output signal at the time of speech recognition. That is, the noise output En
With respect to (z), when En (z) = Xn ₁ (z) z ^-d −Xn ₂ (z) W (z) (1), a parameter that can be adjusted to minimize the power of the error signal e ( Is to determine the coefficient) W of the adaptive filter 3b. Here, it is ideal that Es (z) = Xs ₁ (z) z− ^d −Xs ₂ (z) · W (z) ≒ Xs (z) (2). Note that Xn ₁ (z) and Xn ₂ (z) are noises included in the output signals of the microphones 1 and 2,
Noise sources (noise = xn) from the first, if the transmission characteristics of up to the second microphone 2 and CN1, CN2, be _{Xn 1 (z) = CN1 ·} xn Xn 2 (z) = CN2 · xn , (1) becomes En (z) = (CN1 · z− ^d− CN2 · W (z)) × n (1) ′. Xs ₁ (z) and Xs ₂ (z) are audio signals included in the output signals of the microphones 1 and 2, from the speaker's mouth (speaker voice = xs) to the first and second microphones 1 and 2. if the propagation characteristics of the _{CS1, CS2, xs 1 (z} ) = a _{CS1 · xs xs 2 (z)} = CS2 · xs, (2) expression is Es (z) = (CS1 · z -d -CS2・ W (z)) xs (2) ′

[0007]

Since a vehicle has many noise sources, the coherence of the vehicle interior noise picked up by the microphones 1 and 2 tends to decrease as the microphones 1 and 2 are moved away. . Therefore, there is a problem that the expression (1) becomes larger as the two microphones 1 and 2 are further away, and it is necessary to arrange the microphones 1 and 2 as close as possible. However,
When the two microphones 1 and 2 are arranged as close as possible to each other, it is highly possible that almost the same sound and noise are respectively incident on the two microphones, and the noise is reduced by the adaptive filter coefficient W optimally determined for noise removal. When it is erased, even the voice is erased. On the other hand, (2)
When the adaptive filter coefficient W is determined so as to satisfy the formula, the sound is not damaged but the noise hardly disappears,
There is a problem that the SN ratio is hardly improved.

As described above, it is desirable that two microphones are close to each other in order to suppress noise to the maximum.
On the other hand, in order to minimize sound damage, it is desirable that the two microphones are separated from each other, and they cannot be satisfied simultaneously. For this reason, there is a problem that the conventional microphone system cannot sufficiently improve the S / N ratio of the audio signal. Accordingly, an object of the present invention is to make it possible to improve the S / N ratio of an audio signal in a microphone system (noise reduction system) using two microphones.

[0009]

According to the present invention, there is provided a microphone system for performing adaptive signal processing using output signals of two microphones and outputting a speaker voice signal having an improved SN ratio. This is achieved by arranging two microphones having close proximity to each other and making the angle between the directional direction of the microphone and the utterance direction of the speaker different for each microphone. In this way, 2
Despite placing two microphones close together,
One microphone can pick up the speaker's voice with a high SNR, and the other microphone can pick up the speaker's voice with a low SNR. On the other hand, noise is not greatly reduced in coherence due to different microphone orientations, so that the correlation of the reception noise of each microphone is large, and the difference in the reception sensitivity to the sound of each microphone can be increased,
The SN ratio of the audio signal can be improved. As an example of microphone arrangement, two microphones are approached to the sun visor of the vehicle or the passenger's seat or the driver's seat ceiling, and the angle between the directivity direction of each microphone and the direction of speaker's utterance is made different. And attach it.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS (a) Principle of the present invention In a noise reduction system using two microphones, the correlation between the reception noise of each microphone is increased, and the difference between the reception sensitivities of the microphones for sound is increased. That is ideal. However, there is a trade-off between the “correlation of reception noise” of the two microphones and the “difference in reception sensitivity to voice”, and if one is satisfied by adjusting the distance, the other will not be satisfied.
For example, as the distance between the microphones becomes shorter, the correlation between the reception noises of the two microphones becomes larger, but there is no difference in the reception sensitivity with respect to the voice, and the voice is received in the same manner. Therefore, when adaptive signal processing is performed, noise can be suppressed, but voice is also suppressed at the same time, and as a result, improvement in the SN ratio cannot be expected.

Therefore, according to the present invention, the 2
The two microphones are arranged close to each other, and the angle between the direction of the microphone and the direction of the utterance of the speaker is made different for each microphone. In this way, one microphone picks up the speaker's voice with a high SNR and the other microphone picks up the speaker's voice with a low SNR, despite the two microphones being placed close together. it can. For this reason, the correlation between the reception noises can be increased by the close arrangement of the two microphones, the difference in the reception sensitivity of each microphone with respect to the voice can be increased, and the SN ratio of the voice signal can be improved.

(B) Configuration of Microphone System FIG. 1 is a configuration diagram of a microphone system according to the present invention.
The same parts as those in the system of FIG. 10 are denoted by the same reference numerals. In the figure, 10 is a speaker, for example, a car driver,
11 and 12 are the first having a directional characteristic with respect to the sound receiving sensitivity;
This is the second microphone. The directional characteristic of the microphone has, for example, a unidirectional sensitivity characteristic shown in FIG. That is, when the pointing direction is θ = 0 ⁰ and the sensitivity in the pointing direction is E ₀ , the sensitivity in an arbitrary angle θ direction is expressed by the following equation: E (θ) = E ₀ (1 + cos θ) / 2 The sensitivity decreases as the position shifts.

As an example, the first and second microphones 11 and 12 are mounted on the driver's seat side sun visor 13 at intervals of, for example, 10 cm. The directional direction of the first microphone 11 is provided so as to coincide with the speaker utterance direction (the direction in which the speaker's mouth faces), and the directional direction of the second microphone 12 is oriented in the passenger seat direction. Direction and predetermined angle θ
Is formed. Therefore, the first microphone 11 is more sensitive to the speaker voice than the directivity characteristics of FIG. 2, picks up the speaker voice with a high SN ratio, and
No. 2 has low sensitivity to the speaker's voice and picks up the speaker's voice at a low SN ratio.

[0014] 3 in the adaptive signal processing unit, the output signal x ₂ microphones 12 with the error signal e is input is input as a reference signal, based on the LMS algorithm so that the power of the error signal e is minimized adaptation Perform signal processing.
In the adaptive signal processing unit 3, 3a is an LMS calculation unit and 3b is F
This is an adaptive filter having an IR digital filter configuration. LMS
The arithmetic unit 3a determines the coefficient of the adaptive filter 3b so that the power of the error signal e is minimized by the adaptive signal processing.
The adaptive signal processing unit 3 determines the filter coefficient W of the adaptive filter 3b by adaptive signal processing only during non-speech recognition, does not update the filter coefficient during speech recognition, and uses the filter coefficient W determined during non-speech recognition. Adaptive filter 3
Set to b.

Reference numeral 4 denotes a target response setting unit which receives a signal output from the microphone 11 as a target signal, has a delay characteristic of time d, and has a flat characteristic (gain 1 characteristic) in an audio frequency band. are doing. A subtraction unit 5 subtracts the output signal of the adaptive filter 3b from the target response output from the target response setting unit 4 and outputs an error signal e. According to the microphone arrangement of FIG. 1, in addition to the difference in the sensitivity of the first and second microphones 11 and 12 with respect to the speaker's voice, the sound emission characteristics of human (the sound pressure when the human mouth is used as a sound source is (The characteristic becomes lower as it is deflected to the front of the user), the sound power received by the two microphones 11 and 12 can be made different even though they are arranged close to each other. 1
The correlation between the noises received at 1 and 12 can also be kept high.

(C) Operation At the time of non-speech recognition in which only noise is input to the microphones 11 and 12, the adaptive signal processing unit 3 controls the filter coefficients of the adaptive filter 3b so that the power of the error signal e is minimized by the adaptive signal processing. Determine W. Ideally, the filter coefficient W (Z) is W (z) = CN1 · z− ^d / CN2 (3)

On the other hand, during speech recognition, the adaptive signal processing unit 3 does not update the filter coefficients, but sets the filter coefficient W (Z) determined during the non-speech recognition to the adaptive filter 3b and outputs a speech signal. As a result, the sound signal is expressed by the following equation from equations (2) 'and (3). Becomes If CN1 ≒ CN2 is established by the approach arrangement of microphones, given by equation (4) of the audio signal Es (z) by the following equation Es (z) = (CS1- CS2) · z -d · xs (4) ' Can be Since CS1 ≠ CS2 due to the sensitivity difference between the microphones 1 and 2 and the sound radiation characteristics, the sound signal Es (z) does not become zero. That is, the error signal e at the time of noise input
Even if the adaptive filter coefficient W (Z) is determined so that the power of the audio signal becomes minimum, the audio output Es (z) of the expression (4) does not become zero, and the S / N ratio of the audio signal can be improved. If CN1 ≒ CN2, the magnitude of the audio signal Es (z) mainly depends on the difference between (CS1−CS2), that is, the difference between the sensitivities of the microphones 11 and 12.

(D) Investigation of microphone arrangement and SN ratio improvement amount As described above, in order to improve the SN ratio, while receiving correlated noise with two microphones as much as possible, sound is received with only one microphone as much as possible. It is fundamental to do. Based on this basic principle, the optimal microphone layout was studied. (1) The sun visor of the vehicle and (2) the ceiling on the passenger side of the vehicle were selected as microphone installation locations.

(D-1) Layout of Microphone FIG. 3 (a) is an explanatory view of a layout when a microphone is installed on a sun visor. First and second microphones 11 and 12 are arranged in a sun visor in front of a speaker 10. (Not shown) at an interval of a distance d and fixed so that the directional direction of the first microphone 11 matches the utterance direction,
The angle between the directional direction of the second microphone 12 and the speaker utterance direction was defined as θ. The vertical distance H from the speaker's mouth to each microphone and the distance D from the sun visor are constant and both are about 30.
cm. When examining the amount of improvement in SN ratio, (1)
The positions of the second microphones 11 and 12 were fixed, and the directional direction of the second microphone 12 was changed (see FIG. 4).
(2) The directional directions of the first and second microphones 11 and 12 were fixed, and the position of the second microphone 12 was moved to change the distance d between the microphones (see FIG. 5).

FIG. 3B is an explanatory view of the layout when the microphones 11 and 12 are installed on the passenger side ceiling.
The first and second microphones 11 and 12 are arranged on the front passenger seat side ceiling at an interval of a distance d, and the directional directions of the first and second microphones 11 and 12 are perpendicular to the speaker utterance direction or predetermined. It was arranged to form an angle θ. The vertical distance H and the horizontal distance D from the speaker's mouth to each microphone are constant and both are about 30 cm. When examining the amount of improvement in the SN ratio, (3) the first and second microphones 11 and 12 were set so that their directing directions were perpendicular to the utterance direction, and the position of the second microphone 12 was varied (FIG. 6). (4) The directivity direction of the first microphone 11 is fixed so as to form an angle θ with the direction perpendicular to the utterance direction (installed toward the mouth of the speaker), while the second microphone 12 Was made to be perpendicular to the utterance direction, and the position of the second microphone was varied (see FIG. 7).

(D-2) Results of Study FIGS. 4 to 7 are explanatory diagrams in the case where the improvement in the SN ratio is maximized in the above (1) to (4). In each figure, "Ps"
Is audio power, “Pn” is noise power, “SNR” is SNR,
The “improvement amount” means the SN ratio improvement amount (dB), and the “NR amount” means the noise reduction amount (dB). Further, “before NR” is a value when noise reduction control is not performed (Ps, P at point A in FIG. 1).
n), "After NR" is the value when noise reduction control is performed
(Ps, Pn at point B in FIG. 1). In addition, when considering, "Hachinohe", "Kesenuma", "Gyohashi", "Sapporo",
NR when five voices of "Kitami" were uttered
The Ps, Pn, and SNR before and after the NR were obtained, the SNR improvement amount was obtained from the SNR before and after the NR, and the average value of the SNR improvement amounts was calculated.

(1) FIG. 4 shows a first example of the sun visor.
This is a study result in the case where the positions of the second microphones 11 and 12 are fixed and the angle θ between the directivity direction of the right second microphone 12 and the speaker utterance direction is changed. θ = 15 ⁰ , 30
As a result of examining ⁰ , 45 ⁰ , 60 ⁰ , 90 ⁰ , 120 ⁰ , 180 ⁰ ,
Maximum average SN ratio improvement amount 4.3dB is obtained in theta = 45 ^0.

(2) FIG. 5 shows that the pointing direction of the first microphone 11 is fixed to the speaker uttering direction in the sun visor,
The directional direction of the second microphone 12 is set to 60
The second microphone 12 is fixed at an angle of ⁰
Is a study result in the case where the distance d between the microphones is changed by moving the position. As a result of studying d = 3cm, 6cm, 9cm, 12cm, 15cm and 18cm, the maximum improvement of the average S / N ratio was 4.7dB at d = 9cm.

(3) FIG. 6 shows the first,
This is a study result in a case where the directional directions of the second microphones 11 and 12 are set to be perpendicular to the speaker's utterance direction, and the distance d between the microphones is changed by moving the position of the second microphone 12. As a result of examining d = 2.5 cm, 5 cm, 7.5 cm, d
At 4.5 cm, the maximum average SNR improvement was 4.5 dB.

(4) FIG. 7 shows a state in which the directional direction of the first microphone 11 is fixed on the passenger side ceiling so as to form an angle θ with a direction perpendicular to the speaker's utterance direction, and the second microphone 12 This is a study result in the case where the directivity direction is set to be perpendicular to the speaker utterance direction, and the distance d between the microphones is changed by moving the position of the second microphone. d = 2cm, 4cm, 6cm
As a result, at d = 2 cm, the maximum average SN ratio improvement amount of 4.5 dB was obtained. Microphone as above
By laying out as (1) to (4), SN of 4 to 5 dB
The ratio can be improved. This improvement in the SN ratio makes it possible to greatly improve the recognition rate. Although FIGS. 6 and 7 show examples in which the microphones 11 and 12 are provided on the passenger side ceiling, the microphones may be provided at the same position on the driver side ceiling.

(E) Configuration diagram of another microphone system FIG. 8 is another configuration diagram of a microphone system to which the present invention can be applied, and the same parts as those in FIG. 1 are denoted by the same reference numerals. The difference is that the target response setting unit 4 of FIG. 1 is configured by an adaptive signal processing unit 4 '. In the microphone system shown in FIG. 1, adaptive signal processing is performed only by the adaptive signal processing unit 3 so that the power of the error signal e is minimized. In the microphone system shown in FIG. 8, the adaptive signal processing unit 3 and the adaptive signal processing unit 4 are used. 'Performs adaptive signal processing such that the power of the error signal e is minimized. As described above, the present invention has been described with reference to the embodiments. However, the present invention can be variously modified in accordance with the gist of the present invention described in the claims, and the present invention does not exclude these.

[0027]

As described above, according to the present invention, two microphones having directional characteristics are arranged close to each other, and the angle between the directional direction of the microphone and the utterance direction of the speaker is made different for each microphone. , The S / N ratio of the audio signal output from one microphone can be increased, and the S / N ratio of the audio signal output from the other microphone can be reduced. As a result, the adaptive filter coefficient is determined so that the noise output is minimized. However, the audio output does not become zero, and the S / N ratio of the audio signal can be improved. Further, according to the present invention, each microphone is mounted on a sun visor of a vehicle or on a passenger seat or a driver's seat side ceiling of the vehicle, and each microphone is placed at a relatively short distance with a simple configuration in which the respective directivity directions are different. Despite the placement, one microphone picks up the audio with the highest possible SNR, and the other microphone picks up the lowest possible
Voice can be picked up by the SN ratio, and the SN ratio can be improved.

[Brief description of the drawings]

FIG. 1 is a microphone system of the present invention.

FIG. 2 is an explanatory diagram of directivity characteristics.

FIG. 3 is an explanatory diagram of a layout of a microphone.

FIG. 4 is an explanatory diagram of an SN ratio improvement amount when an angle θ between a directivity direction of a right microphone attached to a sun visor and a speaker utterance direction is changed.

FIG. 5 is fitted with a microphone on the sun visor, 60 ⁰
When moving the right microphone with an angle of
It is a ratio improvement amount explanatory chart.

FIG. 6 shows a case in which each microphone is mounted on the passenger side ceiling so that the directivity direction is perpendicular to the speaker utterance direction, and the distance between the microphones is changed by moving the position of one microphone.
4 is an explanatory diagram of an SN ratio improvement amount.

FIG. 7: A microphone is attached to the front of the passenger side ceiling,
5 is an explanatory diagram of an SN ratio improvement amount when a distance between microphones is changed.

FIG. 8 is a configuration diagram of another microphone system.

FIG. 9 is a diagram illustrating the relationship between the SN ratio and the recognition rate.

FIG. 10 shows the height when two conventional microphones are used.
This is an SN ratio sound receiving system.

FIG. 11 is a characteristic diagram of a target response setting unit.

[Description of Signs] 11, 12 First and second microphones 13 Sun visor 3 Adaptive signal processing unit 3a LMS calculation unit 3b Adaptive filter 4 Target response setting unit 5 Subtraction unit

 ────────────────────────────────────────────────── ─── Continuing on the front page (72) Inventor Shingo Kiuchi 1-8-1, Nishigotanda, Shinagawa-ku, Tokyo Alpine, Inc. F-term (reference) 3D020 BA11 BC02 BC04 BE04 5D015 EE05 KK01 5D020 BB04 BB07 9A001 HH17 KK31

Claims (3)

[Claims] 1. A microphone system for outputting a speaker voice signal having an improved SN ratio by performing adaptive signal processing using output signals of two microphones, wherein two microphones having directional characteristics are arranged close to each other, A microphone system characterized in that the angle formed by the microphone's directional direction and the speaker's utterance direction differs for each microphone. 2. The microphone system according to claim 1, wherein each of the microphones is attached to a sun visor of a vehicle. 3. The microphone system according to claim 1, wherein each of the microphones is mounted on a driver's seat or a passenger side ceiling of a vehicle.