JP2016048872A - Sound collection device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a sound collection device which makes a target sound clear by collecting the target sound and an external noise sound while utilizing a plurality of microphones in a predetermined positional relation.SOLUTION: A sound collection device 100 includes a first microphone 110, two second microphones 120-1 and 120-2 and a signal processing part. The first microphone is a head-set type and disposed in the vicinity of the mouth of a wearer for collecting the target sound that is a voice uttered by the wearer. The second microphones are disposed at position away from the first microphone 110 in a head-set style for collecting the external noise sound. The signal processing part uses a sound collection signal of the first microphone 110 and sound collection signals of the second microphones 120-1 and 120-2 to generate an output signal in which the target sound is emphasized and the external noise sound is suppressed. The first microphone 110 has unidirectionality in a direction of the mouth of the wearer, and the two second microphones 120-1 and 120-2 are disposed in the vicinity of both ears of the wearer and have directionality outwards.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a headset type sound collecting device.

  Non-Patent Document 1 is known as noise canceling headphones. In Non-Patent Document 1, noise is collected by a microphone built in the headphone, an opposite phase sound having an effect of canceling the noise is generated, and reproduced together with an audio signal. Thereby, the user can hear a voice signal more clearly.

“Product Information> Headphones> Lineup> MDR-1RNC> Product Features” [online], Sony Corporation, Sony Marketing (Japan) Inc., [searched July 24, 2014], Internet <URL: http: //www.sony.jp/headphone/products/MDR-1RNC/feature_1.html>

  However, the conventional technology assumes that noise around a headphone is canceled with respect to the reproduced sound / sound signal when the sound / sound signal recorded or recorded sound is reproduced with headphones in another environment. Therefore, it is not assumed that the noise around the headset is canceled with respect to the collected sound signal collected by the microphone of the headset. A headphone is a device that converts electrical signals output from a playback device or receiver into sound waves (audible sound) using a sounding body (such as a speaker) close to the ear (the eardrum). A generic term for microphones (sound pickup devices) worn on the head.

  The present invention collects a target sound and an external noise sound using a plurality of microphones in a predetermined positional relationship, and performs a signal processing so that the target sound is clearer than before. An object is to provide an apparatus.

  In order to solve the above problems, according to one aspect of the present invention, the sound collection device is a headset type. The sound collection device is arranged near the mouth of the wearer of the sound collection device, and a first microphone for collecting a target sound, which is a sound emitted by the wearer, (I) Target sound by using two second microphones for collecting external noise sound, a collected sound signal of the first microphone, and a collected sound signal of the second microphone. And / or (ii) a signal processing unit that generates an output signal in which external noise sound is suppressed. The first microphone has a single directivity with respect to the direction of the wearer's mouth, and the two second microphones are respectively disposed in the vicinity of both ears of the wearer and have directivity with respect to the outward direction. Have.

  According to the present invention, there is an effect that signal processing can be performed on a sound collection signal so that the target sound is clearer than a conventional headset type sound collection device.

The figure which shows the structure of the sound collection device which concerns on 1st embodiment. The functional block diagram of the sound collection device which concerns on 1st embodiment. The figure which shows the example of the processing flow of the sound collection device which concerns on 1st embodiment. The block diagram of the filter estimation part which concerns on 2nd embodiment. The figure which shows the example of the processing flow of the filter estimation part which concerns on 2nd embodiment. The figure for demonstrating the example of gain shaping. The block diagram of the sound collection device which concerns on 3rd embodiment. The figure which shows the example of the processing flow of the sound collection device which concerns on 3rd embodiment. The block diagram of the filter estimation part which concerns on the modification of 3rd embodiment. The figure which shows the example of the processing flow of the filter estimation part which concerns on the modification of 3rd embodiment.

  Hereinafter, embodiments of the present invention will be described. In the drawings used for the following description, constituent parts having the same function and steps for performing the same process are denoted by the same reference numerals, and redundant description is omitted. In the following explanation, the symbol “^” etc. used in the text should be described immediately above the character immediately after it, but it is described immediately before the character due to restrictions on the text notation. In the formula, these symbols are written in their original positions. Further, the processing performed for each element of a vector or matrix is applied to all elements of the vector or matrix unless otherwise specified.

<First embodiment>
FIG. 1 is a diagram illustrating a configuration of a sound collection device 100 according to the first embodiment, FIG. 2 is a functional block diagram thereof, and FIG. 3 is a diagram illustrating a processing flow thereof.

  The sound collection device 100 includes a first microphone 110, second microphones 120-1 and 120-2, and a signal processing unit 140.

  The sound collection device 100 (i) emphasizes the target sound from the sound collection signals collected by the first microphone 110 and the second microphones 120-1 and 120-2 and / or (ii) external noise sound. An output signal z in which is suppressed is generated and output.

<Headset shape, first microphone and second microphone>
The sound collection device 100 is a headset type. As described above, a headset is a general term for a microphone (sound collecting device) worn on the head, and the structure of the headset is that the fixed band 101 passes through the top of the wearer 9 via the ear pad 102 and the like. 9 supported by both ears of the wearer 9, supported by the ears and the back of the wearer 9 via the ear pad 102 and the fixed band 101, and supported by the wearer 9 via the fixed band 101 and the temple support portion not shown. Some are supported by the back of the head and temples. In addition, any form that can be mounted on the head and that can mount the following first microphone and second microphone may be used.

  The first microphone is a microphone for picking up a target sound that is a sound emitted by the wearer 9. In the present embodiment, there is one first microphone (first microphone 110), which is disposed in the vicinity of the mouth of the wearer 9, and has a single directivity with respect to the direction of the mouth of the wearer 9 (FIG. 1). reference). With such a configuration, a target sound that is a sound emitted by the wearer 9 is collected.

  The second microphone is a microphone for collecting external noise sound. In the present embodiment, there are two second microphones (second microphones 120-1 and 120-2), which are arranged in the vicinity of both ears of the wearer 9 and have directivity toward the outside (FIG. 1). reference). With such a configuration, an external noise sound is collected.

  In the case of the headset, it is considered that the sound to be collected is limited to the sound emitted from the mouth of the wearer 9. Therefore, in the present embodiment, the first microphone is arranged near the mouth of the wearer 9, and the first microphone is selected to have directivity in order to emphasize the sound of the mouth as much as possible. In addition, the second microphone for emphasizing the external noise sound and picking up the sound is arranged at a position away from the first microphone, and has directivity outward.

<Processing content of signal processing unit 140>
The signal processing unit 140 receives the collected sound signal of the first microphone and the collected sound signal of the second microphone from the first microphone and the second microphone, and using these values, (i) emphasizes the target sound, or / And (ii) An output signal in which the external noise sound is suppressed is generated and output as an output value of the sound collection device 100.

  For example, the signal processing unit 140 includes square units 141 and 142, a filter estimation unit 143, and a second filtering unit 144.

The signal processing unit 140 collects the sound collected signal X _{0 obtained} by converting the time domain signal collected by the first microphone 110 into a frequency domain signal, and the second microphones 120-1 and 120-2, respectively. The frequency domain sound pickup signals X ₁ and X ₂ obtained by converting the time domain signal thus converted into a frequency domain signal are received, and an output signal z is output.

<Square part 141>
The square unit 141 receives the collected sound signals X ₁ and X ₂ of the second microphones 120-1 and 120-2, adds these values and squares them, or the square sum of these values, or these A value obtained by weighting and adding a value obtained by squaring the value of is calculated (S141) and output. Note that ^ φ _N is an estimate of the power spectral density of the noise area defined by the following equation. For example,

Or

Or

And However, K is the number of second microphones (therefore, K = 2 in this embodiment), ω is a frequency, τ is a frame index, and g _k (ω) is a preset constant (weight). Details of the noise area will be described later.

  In the case of (1), there is an advantage that the power spectrum density of the noise area can be estimated more accurately than using (2).

  In the case of (2), the error in estimating the power spectral density of the noise area is larger than in (1), but the addition process can be performed by an analog circuit, and the number of second microphones is 2 or more. Also, there is an advantage that a hardware configuration can be achieved with one AD converter, and an inexpensive hardware configuration can be achieved.

  In the case of (3), there is an advantage that the weight for the second microphone containing more external noise sounds than the target sound can be set larger, and the power spectral density of the noise area can be estimated more accurately. For example, by increasing the weight of the second microphone farthest from the mouth, it is possible to increase the weight of the second microphone with a small amount of target sound and increase the estimation accuracy of the power spectrum density in the noise area.

<Square part 142>
The square unit 142 receives the sound pickup signal X ₀ of the first microphone 110, calculates a value ^ φ _S (ω, τ) = | X ₀ (ω, τ) | ² obtained by squaring this value (S142), Output. Note that ^ φ _S is an estimated value of the power spectral density of the target area. Details of the target area will be described later.

<Filter estimation unit 143>
The filter estimation unit 143 receives ^ φ _N (ω, τ) and ^ φ _S (ω, τ), estimates the filter G that suppresses the external noise sound (S143), and outputs it.

  For example, a post filter design method based on Reference 1 will be described.

(Reference 1) Y. Hioka et al., “Underdetermined sound source separation using power spectrum density estimated by combination of directivity gain,” IEEE Trans. Audio, Speech, Language Proc., 21, 1240-1250, 2013.
Reference 1 proposes a method of designing a post filter based on the power spectral density (PSD) of each area estimated using a plurality of beamforming. Hereinafter, this method is referred to as an LPSD method (Local PSD-based post-filter design). The processing flow of the LPSD method will be described with reference to FIG.

  For example, when designing a post filter based on the Wiener method, the filter G (ω, τ) is calculated as follows.

Here, φ _S (ω, τ) is a target area (in the case where the periphery of the microphone is divided into a plurality of areas in advance, it is an area including a sound source to be picked up from the plurality of areas. Included area: In the case of a headset, it represents the power spectral density of the wearer's mouth, and φ _N (ω, τ) is a noise area (an area containing a sound source that emits external noise sound) In the case of the headset, the power spectral density of the above-mentioned plurality of areas that do not include the wearer's mouth except the target area is represented. Here, the power spectrum density of a certain area means the power spectrum density of sound coming from that area. That is, for example, the power spectral density of the target area is the power spectral density of sound coming from the target area, and the power spectral density of the noise area is the power spectral density of sound coming from the noise area.

As will be described later, in the case of a headset, the first microphone and the second microphone are in a predetermined positional relationship (the second microphone is arranged at a position separated from the first microphone in the form of the headset). Although there is a positional relationship is not the same. In other words, the positional relationship changes for each wearer (according to the size of the head and the position of the mouth), every time it is worn, and every time. Therefore, the difference in phase and amplitude generated between the microphones changes. Therefore, in beam forming that uses the difference in phase and amplitude generated between microphones, the performance changes sensitively with respect to the positional relationship between the microphones. It is not suitable as a means for estimating the noise level. Therefore, the first microphone is a microphone having directivity in the direction of the wearer's mouth. Furthermore, the target sound is mainly included in the collected sound signal by the first microphone by bringing the first microphone close to the mouth. On the other hand, the second microphone is a microphone having directivity with respect to the outward direction. Furthermore, in order to robustly estimate the level of the external noise sound relative to the position of the microphone, the second microphone is arranged at a position away from the first microphone arranged in the vicinity of the mouth, so that sound collection by the second microphone is performed. The signal mainly includes external noise. Thus, the filter estimating unit 143 estimates the power spectral density of the target area ^ phi _S is collected sound signal X ₀ of the first microphone (omega, tau) with _{^ φ S = | X 0 (} ω, τ ) | ² , and the estimated value ^ φ _N of the power spectral density in the noise area is obtained by, for example, equations (1) to (3) using the sound pickup signal X _k (ω, τ) of the second microphone.

  For example, the filter estimation unit 143 estimates the filter G (ω, τ) by the following equation.

<Second filtering unit 144>
The second filtering unit 144 receives the filter G, using the filter G, performs filtering on the sound pickup signal X ₀ (S144). In order to suppress the external noise sound included in X ₀ (ω, τ), the post filter G (ω, τ) is multiplied.

  Finally, the output signal z is obtained by performing inverse fast Fourier transform (IFFT) on Z (ω, τ).

<Effect>
With such a configuration, signal processing can be performed so that the target sound is clearer than the conventional headset type sound pickup device. In particular, it is possible to improve the easiness of hearing the target sound collected under high noise, and it is possible to make a call and recognize voice under high noise.

  In Non-Patent Document 1, since it is assumed that noise cancellation is performed on the sound signal of the headphones, it is not necessary to collect the sound emitted by the wearer. For this reason, there is no microphone for picking up sound emitted by the wearer. Even if the headphones of Non-Patent Document 1 and a conventional headset (microphone for collecting sound emitted by the wearer) are combined, the sound that is subject to noise cancellation is an audio signal of the headphones, It is not a sound pickup signal picked up by a microphone for picking up sound emitted by the wearer. Therefore, there is no problem in the conventional technology, and the positional relationship between the microphone for collecting the sound emitted by the wearer and the microphone for collecting the external noise sound has not been studied.

  A method of arranging a plurality of microphones at the position of the first microphone (near the mouth) so that the positional relationship does not change and performing beam forming is also conceivable, but there is a problem that the amount of calculation of the filtering process increases. Furthermore, a space is limited at the tip of the headset such as a flexible pipe, which is not suitable for arranging a plurality of microphones.

  In this embodiment, by designing a device that can collect the target sound and other external noise sounds as much as possible by using a directional microphone, etc., the external noise sound can be obtained without beam forming. It was set as the structure which can perform level estimation.

<Modification>
You may combine the conventional noise canceling headphones and the headset of this embodiment. In that case, the collected sound signal of the second microphone may be used to perform noise cancellation on the sound signal of the headphones.

The first microphone is not limited to the configuration of the first embodiment as long as it can collect the target sound and not the external noise sound as much as possible. For example, the sound collection device 100 may include a plurality of first microphones. Further, when the first microphone is brought close to the mouth of the wearer 9 so that the target sound is mainly included in the collected sound signal by the first microphone, the first microphone is not necessarily unidirectional. It does not have to be, and any directivity is acceptable. Further, it is not necessarily arranged near the mouth of the wearer 9. For example, it may be arranged near the ear, have super directivity, and collect the target sound. Even in such a configuration, since the target sound is mainly included in the sound pickup signal X ₀ (ω, τ) by the first microphone, the estimated value of the power spectral density of the target area ^ φ _S (ω , τ) _{is, ^ φ S (ω, τ} ) = | X 0 (ω, τ) | can be estimated at ^2.

The second microphone is not limited to the configuration of the first embodiment as long as it can collect the target sound and collect the external noise as much as possible. For example, the sound collection device 100 may include only one second microphone or may include three or more second microphones. In addition, the second microphone does not necessarily have to be disposed near both ears of the wearer 9 and may be disposed at a position where the target sound is not collected and the external noise sound can be collected as much as possible. . For example, A to F in FIG. 1 (A is the vicinity of the left cheek, B is the left occipital region (1) or the left occipital region, C is the left occipital region (2) or the occipital region, D is the occipital region or the occipital region, E is the right occipital region) (2) or right top of the head, F is located at the position of the right back of the head (1) or right side of the head) and has directivity so as not to pick up the target sound as much as possible, and picks up external noise sound. May be. Note that the second microphone and the first microphone 110 are arranged in a physically separated position in the form of a headset, so that the target sound is not collected as much as possible and the external noise sound is collected. It can be made easier. In particular, when the second microphones are arranged in the vicinity of both ears of the wearer 9 as in the first embodiment, the target sound is not collected and the external noise sound is easily collected. In such a positional relationship, when the target sound is not picked up and the external noise sound can be picked up easily, the second microphone may not have directivity with respect to the outward direction, Any directivity is acceptable. Even in such a configuration, an external noise sound is mainly included in the sound pickup signal X _k (ω, τ) from the second microphone, so an estimated value ^ φ _N ( (ω, τ) can be estimated by equations (1) to (3).

  The filter estimated by the filter estimation unit 143 may be any filter as long as it (i) emphasizes the target sound and / or (ii) suppresses the external noise sound. It is not limited to the filter.

<Second embodiment>
A description will be given centering on differences from the first embodiment. The process in the filter estimation part 143 differs from 1st embodiment. In the present embodiment, “external noise sound” is also referred to as “noise”.

The LPSD method has formulated the problem on the assumption that the target sound and interference noise are mixed. However, practical problems often include not only coherent interference noise but also stationary noise with high incoherence (air conditioning noise, microphone internal noise, etc.). In this case, estimation errors of φ _S (ω, τ) and φ _N (ω, τ) become large, and noise suppression performance may be deteriorated.

  In the filter estimation unit 143 described below, the post-filter is robustly estimated against various noise environments by extending the LPSD method. Specifically, the estimation error of the ratio between the power of the target sound and the power of other noise is reduced by estimating the power spectral density by dividing each noise type.

  FIG. 4 shows a block diagram of an example of the filter estimation unit 143.

  As shown in FIG. 4, the filter estimation unit 143 includes a first steady / unsteady component extraction unit 143A, a second steady / unsteady component extraction unit 143B, a multi-noise corresponding gain calculation unit 143C, and a time frequency average For example, a conversion unit 143D and a gain shaping unit 143E are provided.

  FIG. 5 shows each step of signal processing realized by the filter estimation unit 143 of the sound collection device, for example.

  Details of the filter estimation unit 143 and the method embodiment of the sound collection device will be described below. The basic signal processing framework, definition of words, and the like are the same as those described in the background art and the first embodiment. Therefore, these overlapping explanations are omitted.

  In the present embodiment, it is assumed that the target sound is collected by the first microphone 110 and the external noise sound is collected by the second microphones 120-1 and 120-2.

<First Steady / Unsteady Component Extraction Unit 143A>
For example, ^ φ _S (ω, τ) defined by the following equation has a nonstationary component ^ φ _S ^(A) (ω, τ) derived from the sound arriving from the target area and a steady state derived from incoherent noise. The component ^ φ _S ^(B) (ω, τ) is included.

  K represents the number of second microphones. Here, there are two types of noise, interference noise and incoherent noise. Interference noise is noise generated from a noise source arranged in a noise area. Incoherent noise is not limited to noise emitted from the target area and the noise area, but is noise that is emitted from the noise area and places other than these areas and exists constantly.

Therefore, the first stationary / unsteady component extraction unit 143A determines the unsteady component ^ φ _S ^(A) (derived from the sound arriving from the target area from the power spectral density ^ φ _S (ω, τ) of the target area. The stationary component ^ φ _S ^(B) (ω, τ) derived from ω, τ) and incoherent noise is extracted by time averaging (S143A).

There are various non-stationary components ^ φ _S ^(A) (ω, τ) derived from the sound coming from the extracted target area and stationary components ^ φ _S ^(B) (ω, τ) derived from incoherent noise. It is output to the noise corresponding gain calculation unit 143C.

For example, the first stationary / unsteady component extraction unit 143A performs an exponential moving average process as shown in Equation (11) and Equation (12), so that ^ φ _S (ω, τ) to ^ φ _S ^{(B )} Calculate (ω, τ).

Here, α _S is a smoothing coefficient, which is a predetermined positive real number. For example, 0 <α _S <1. Further, the time constant may be set to about 150 ms. Υ _S is a set of frames index for a specific section. For example, the specific section is set to be about 3 to 4 seconds.

Then, the first constant / non-stationary component extracting section 143A, as in the equation _{(13), ^ φ S (} ω, τ) from ^{_{^ φ S (B) (ω}} , τ) by subtracting a ^ phi _S ^(A) (ω, τ) is calculated.

Here, β _S (ω) is a weighting coefficient, which is a predetermined positive real number. β _S (ω) is set to a real number of about 1 to 3, for example.

Note that ^ φ _S ^(A) (ω, τ) may be floored so as to satisfy the condition of ^ φ _S ^(A) (ω, τ) ≧ 0. This flooring process is performed by the first stationary / unsteady component extraction unit 143A, for example.

<Second Steady / Unsteady Component Extraction Unit 143B>
For example, ^ φ _N (ω, τ) defined by equation (10) includes non-stationary components derived from interference noise ^ φ _N ^(A) (ω, τ) and stationary components derived from incoherent noise ^ φ _N ^(B) (ω, τ) is included.

Therefore, the second stationary / unsteady component extraction unit 143B receives the power spectral density ^ φ _N (ω, τ) of the noise area as an input, and generates interference noise from the power spectral density ^ φ _N (ω, τ) of the noise area. The non-stationary component ^ φ _N ^(A) (ω, τ) derived from 及 び and the stationary component ^ φ _N ^(B) (ω, τ) derived from incoherent noise are extracted by time averaging (S143B).

The non-stationary component ^ φ _N ^(A) (ω, τ) derived from the extracted interference noise and the stationary component ^ φ _N ^(B) (ω, τ) derived from incoherent noise are It is output to the calculation unit 143C.

For example, the second stationary / unsteady component extraction unit 143B performs an exponential moving average process as shown in Equation (14) and Equation (15), so that ^ φ _N (ω, τ) to ^ φ _N ^(B) Calculate (ω, τ).

Here, α _N is a smoothing coefficient, which is a predetermined positive real number. For example, 0 <α _N <1. Further, the time constant may be set to about 150 ms. Υ _N is a set of frames index for a specific section. For example, the specific section is set to be about 3 to 4 seconds.

Then, the second constant / non-stationary component extracting section 143B, as in the equation _{(16), ^ φ N (} ω, τ) from ^{_{^ φ N (B) (ω}} , τ) by subtracting a ^ phi _N ^(A) Calculate (ω, τ).

Here, β _N (ω) is a weighting coefficient, which is a predetermined positive real number. β _N (ω) is set to a real number of about 1 to 3, for example.

Note that ^ φ _N ^(A) (ω, τ) may be floored so as to satisfy the condition of ^ φ _N ^(A) (ω, τ) ≧ 0. This flooring process is performed by, for example, the second steady / unsteady component extraction unit 143B.

α _N may be the same as or different from α _S. Υ _N may be the same as or different from Υ _S. β _N (ω) may be the same as or different from β _S (ω).

<Gain calculation unit 143C for various noises>
The various noise corresponding gain calculation unit 143C is configured such that the non-stationary component ^ φ _S ^(A) (ω, τ) derived from the sound arriving from the target area and the stationary component ^ φ _S ^(B) derived from incoherent noise. (ω, τ), non-stationary component derived from interference noise ^ φ _N ^(A) (ω, τ), and stationary component derived from incoherent noise ^ φ _N ^(B) (ω, τ) These are used to calculate the filter ~ G (ω, τ) that emphasizes the steady component of the sound coming from the target area (S143C).

  The calculated filter ~ G (ω, τ) is output to the time frequency averaging unit 143D.

  Since the power spectral density is estimated for each type of noise (in other words, for each type of noise such as incoherent noise and coherent noise), the various noise corresponding gain calculation unit 143C uses, for example, the following equation (17). Calculate the defined post filter ~ G (ω, τ).

If there is a difference between the behavior of the value of ^ φ _S ^(B) (ω, τ) and the behavior of ^ φ _N ^(B) (ω, τ), and the assumption of incoherence is broken, The various noise corresponding gain calculation unit 143C may calculate a post filter ~ G (ω, τ) defined by the following equation (18).

<Time frequency averaging unit 143D>
The time frequency averaging unit 143D receives the filter ~ G (ω, τ), and performs a smoothing process on the filter ~ G (ω, τ) in at least one of the time direction and the frequency direction (S143D).

  The smoothed filter˜G (ω, τ) is output to the gain shaping unit 143E.

In the case of performing smoothing in the time direction, τ ₀ and τ ₁ are integers greater than or equal to 0, and the time frequency averaging unit 143D is, for example, a filter in the time direction of filters to G (ω, τ). What is necessary is just to perform an addition average about ~ G (ω, τ-τ ₀ ), ... ~ G (ω, τ + τ ₁ ). The time-frequency averaging unit 143D may perform weighted addition for ~ G (ω, τ-τ ₀ ), ... ~ G (ω, τ + τ ₁ ).

Further, when performing smoothing in the frequency direction, ω ₀ and ω ₁ are set to real numbers of 0 or more, and the time frequency averaging unit 143D performs, for example, a filter in the vicinity of the frequency direction of filters to G (ω, τ). What is necessary is just to perform an addition average for ~ G (ω−ω ₀ , τ),... ~ G (ω + ω ₁ , τ). The time frequency averaging unit 143D may perform weighted addition for ~ G (ω-ω ₀ , τ), ... ~ G (ω + ω ₁ , τ).

<Gain shaping unit 143E>
The gain shaping unit 143E receives the smoothed filter ~ G (ω, τ) and performs the gain shaping on the smoothed filter ~ G (ω, τ), thereby obtaining the filter G ( (ω, τ) is generated and output (S143E). The gain shaping unit 143E generates, for example, a filter G (ω, τ) defined by the following equation (19).

  Here, γ is a weighting factor, which is a positive real number. For example, γ may be set to about 1 to 1.3.

  The gain shaping unit 143E may perform a flooring process on the filter G (ω, τ) so as to satisfy A ≦ G (ω, τ) ≦ 1. A is a real number from 0 to 0.3, usually about 0.1. If G (ω, τ) is larger than 1, there is a possibility of overemphasis, and if G (ω, τ) is too small, musical noise may be generated. By performing an appropriate flooring process, it is possible to prevent this enhancement and the generation of musical noise.

  Consider a function f whose domain and range are real numbers. For example, the function f is a non-decreasing function. Gain shaping means an operation for obtaining an output value when ~ G (ω, τ) before gain shaping is input to the function f. In other words, the output value when ~ G (ω, τ) is input to the function f is G (ω, τ). An example of the function f is Expression (19). The function f according to the equation (19) is f (x) = γ (x−0.5) +0.5.

  Another example of another function f will be described with reference to FIG. In FIG. 6, the index is omitted. That is, G in FIG. 6 means G (ω, τ), and ~ G means ~ G (ω, τ). First, in this example, as shown in FIGS. 6A to 6B, the slope of the graph of the function f is changed. Then, as shown in FIGS. 6B to 6C, flooring processing is performed so as to satisfy 0 ≦ G (ω, τ) ≦ 1. The function specified by the graph indicated by the bold line in FIG. 6C is another example of the function f.

  The graph of the function f is not limited to that shown in FIG. For example, in FIG. 6C, the graph of the function f is composed of a straight line, but the graph of the function f may be composed of a curve. For example, the function f may be a function obtained by performing a flooring process on a hyperbolic tangent function.

<Effect>
With such a configuration, the same effect as that of the first embodiment can be obtained. Furthermore, according to the filter estimation unit 143, it is possible to design a post filter for suppressing noise robustly in an environment where noise having various properties exists. In addition, such a post filter can be designed by processing with real-time characteristics.

<Modification>
The processing of the time frequency averaging unit 143D and the gain shaping unit 143E is performed to suppress so-called musical noise. The processing of the time frequency averaging unit 143D and the gain shaping unit 143E may not be performed.

The calculation of ^ φ _S ^(B) (ω, τ) and ^ φ _S ^(A) (ω, τ) by the exponential moving average process is an example of the process of the first steady / unsteady component extraction unit 143A. The first steady / unsteady component extraction unit 143A may extract ^ φ _S ^(B) (ω, τ) and ^ φ _S ^(A) (ω, τ) by other processing.

Similarly, calculation of ^ φ _N ^(B) (ω, τ) and ^ φ _N ^(A) (ω, τ) by exponential moving average processing is an example of processing of the second stationary / unsteady component extraction unit 143B. is there. The second steady / unsteady component extraction unit 143B may extract ^ φ _N ^(B) (ω, τ) and ^ φ _N ^(A) (ω, τ) by other processing.

<Third embodiment>
A description will be given centering on differences from the first embodiment.

  Although the first microphone is close to the wearer 9, the external noise sound is also mixed in the sound pickup signal of the first microphone. In addition, although the second microphone is disposed at a position away from the first microphone, the target sound is mixed in the collected sound signal of the second microphone. In the present embodiment, the accuracy of the filter G is improved by correcting the mixed amount.

Therefore, in this embodiment, the estimated value ^ φ _S (ω, τ) of the power spectral density in the target area and the estimated value ^ φ _N (ω, τ) of the power spectral density in the noise area are corrected. The correction formula is as follows.
^ φ ' _S (ω, τ) = ^ φ _S (ω, τ) -α ^ φ _N (ω, τ) (21)
^ φ ' _N (ω, τ) = ^ φ _N (ω, τ) -γ ^ φ _S (ω, τ) (22)
The filter estimation unit 143 replaces ^ φ _S (ω, τ) and ^ φ _N (ω, τ) in Equation (5) with ^ φ ′ _S (ω, τ) and ^ φ ′ _N (ω, The post filter G (ω, τ) may be calculated using τ).

  Note that α and γ may be set in advance or may be set adaptively in consideration of the arrangement and directivity characteristics of the first microphone and the second microphone.

  In the case of setting in advance, the functional block diagram of the signal processing unit 140 is represented in FIG. 2, and the filter estimation unit 143 may store α and γ in advance.

  FIG. 7 is a functional block diagram of the signal processing unit 140 when α and γ are adaptively set, and FIG. 8 shows an example of the processing flow. The signal processing unit 140 includes a level ratio estimation unit 145.

<Level Ratio Estimator 145>
The level ratio estimation unit 145 receives the estimated value ^ φ _S (ω, τ) of the power spectral density of the target area and the estimated value ^ φ _N (ω, τ) of the power spectral density of the noise area, and uses these values. Then, α and γ are obtained (S145) and output to the filter estimation unit 143.

α is a level difference between the first microphone and the second microphone of the noise component. Therefore, in a section (noise section) where there is no target sound and only noise exists (level of the first microphone) / (level of the second microphone) ) Is calculated. First,
α = ^ φ _S (ω, τ) / ^ φ _N (ω, τ) (23)
Calculate Since the target sound enters the first microphone largely and enters the second microphone smallly, by observing the level difference between the microphones (level of the first microphone) / (level of the second microphone), the section where the target sound exists Then, it is possible to identify a noise-only section (noise section) in which the target sound does not exist. When the level difference (the level of the first microphone) / (the level of the second microphone) is equal to or less than a preset threshold value, it is determined that it is a noise interval, and α at that time is output. For example, the target sound collected by the first microphone is expected to be about 10 to 20 dB higher than the second microphone. Therefore, the threshold value set in advance is preferably set between 1 and 10.

Since γ is a level difference between the second microphone and the first microphone of the target sound, it can be obtained by calculating (level of the second microphone) / (level of the first microphone) in the speech section. First,
γ = ^ φ _N (ω, τ) / ^ φ _S (ω, τ) (24)
Calculate Since the target sound enters the first microphone largely and enters the second microphone smallly, the speech interval can be detected by observing the level difference between the microphones (second microphone level) / (first microphone level). Can do. When the level difference (the level of the second microphone) / (the level of the first microphone) is equal to or less than a preset threshold value, the speech section is determined to be output, and γ at that time is output. For example, the level difference between the first microphone and the second microphone is expected to be about 10 to 20 dB in the utterance period. Therefore, the threshold value set in advance is preferably set between 1 and 10.

  The level ratio estimation unit 145 obtains α and γ at regular intervals, so that the filter estimation unit 143 can adaptively set α and γ.

<Effect>
With this configuration, the same effect as that of the first embodiment can be obtained. Furthermore, the accuracy of the filter G can be improved by taking into account the mixing of external noise sound in the first microphone and the mixing of target sound in the second microphone.

<Modification>
You may combine this embodiment and 2nd embodiment.

Before the processing of the filter estimation unit 143 of the second embodiment, ^ φ ' _S (ω, τ) and ^ φ' _N (ω, τ) are calculated by the method of the third embodiment, and ^ φ ' _S The processing of the filter estimation unit 143 of the second embodiment may be performed using (ω, τ) and ^ φ ′ _N (ω, τ), but in this modification, it is derived from the sound coming from the target area The unsteady component ^ φ _S ^(A) (ω, τ) and the unsteady component ^ φ _N ^(A) (ω, τ) derived from environmental noise are corrected. The correction formula is as follows.
^ φ ' _S ^(A) (ω, τ) = ^ φ _S ^(A) (ω, τ) -β ^ φ _N ^(A) (ω, τ) (25)
^ φ ' _N ^(A) (ω, τ) = ^ φ _N ^(A) (ω, τ) -κ ^ φ _S ^(A) (ω, τ) (26)
The variable noise corresponding gain calculation unit 143C replaces ^ φ _S ^(A) (ω, τ) and ^ φ _N ^(A) (ω, τ) in Equation (17) or Equation (18) with ^ φ The post filter ~ G (ω, τ) may be calculated using ' _S ^(A) (ω, τ) and ^ φ' _N ^(A) (ω, τ).

  Note that β and κ may be set in advance or adaptively in consideration of the arrangement and directivity characteristics of the first microphone and the second microphone.

  In the case of setting in advance, the functional block diagram of the filter estimation unit 143 is shown in FIG. 4, and the various noise corresponding gain calculation unit 143C may store β and κ in advance.

  FIG. 9 is a functional block diagram of the filter estimation unit 143 when β and κ are adaptively set, and FIG. 10 shows an example of the processing flow. The filter estimation unit 143 includes a level ratio estimation unit 143F.

The level ratio estimation unit 143F arrives from the target area instead of the estimated value ^ φ _S (ω, τ) of the power spectral density of the target area and the estimated value ^ φ _N (ω, τ) of the power spectral density of the noise area. Using the non-stationary component ^ φ _S ^(A) (ω, τ) derived from the sound and the non-stationary component ^ φ _N ^(A) (ω, τ) derived from the environmental noise, Instead of γ and γ, β and κ are obtained and output (S143F). Therefore,
β = ^ φ _S ^(A) (ω, τ) / ^ φ _N ^(A) (ω, τ)
κ = ^ φ _N ^(A) (ω, τ) / ^ φ _S ^(A) (ω, τ)
And are output when β and κ are each equal to or less than a preset threshold value.

  The present invention is not limited to the above-described embodiments and modifications. For example, the various processes described above are not only executed in time series according to the description, but may also be executed in parallel or individually as required by the processing capability of the apparatus that executes the processes. In addition, it can change suitably in the range which does not deviate from the meaning of this invention.

<Program and recording medium>
In addition, various processing functions in each device described in the above embodiments and modifications may be realized by a computer. In that case, the processing contents of the functions that each device should have are described by a program. Then, by executing this program on a computer, various processing functions in each of the above devices are realized on the computer.

  The program describing the processing contents can be recorded on a computer-readable recording medium. As the computer-readable recording medium, for example, any recording medium such as a magnetic recording device, an optical disk, a magneto-optical recording medium, and a semiconductor memory may be used.

  The program is distributed by selling, transferring, or lending a portable recording medium such as a DVD or CD-ROM in which the program is recorded. Further, the program may be distributed by storing the program in a storage device of the server computer and transferring the program from the server computer to another computer via a network.

  A computer that executes such a program first stores, for example, a program recorded on a portable recording medium or a program transferred from a server computer in its storage unit. When executing the process, this computer reads the program stored in its own storage unit and executes the process according to the read program. As another embodiment of this program, a computer may read a program directly from a portable recording medium and execute processing according to the program. Further, each time a program is transferred from the server computer to the computer, processing according to the received program may be executed sequentially. Also, the program is not transferred from the server computer to the computer, and the above-described processing is executed by a so-called ASP (Application Service Provider) type service that realizes the processing function only by the execution instruction and result acquisition. It is good. Note that the program includes information provided for processing by the electronic computer and equivalent to the program (data that is not a direct command to the computer but has a property that defines the processing of the computer).

  In addition, although each device is configured by executing a predetermined program on a computer, at least a part of these processing contents may be realized by hardware.

  Speech recognition has been commonly used as a command input for smartphones. Under noisy conditions such as in cars and factories, there is a high demand for hands-free operation of devices and calls with remote locations.

  The present invention can be used, for example, in such a case.

Claims (2)

A headset type sound collecting device,
A first microphone arranged in the vicinity of the mouth of the wearer of the sound collection device for collecting a target sound, which is a sound emitted by the wearer;
The first microphone is arranged at a position separated in the form of a headset, and two second microphones for collecting external noise sound,
Signal processing for generating an output signal in which (i) the target sound is emphasized and / or (ii) the external noise sound is suppressed by using the collected sound signal of the first microphone and the collected sound signal of the second microphone Including
The first microphone has unidirectionality with respect to the direction of the wearer's mouth,
Each of the two second microphones is disposed in the vicinity of both ears of the wearer and has directivity toward the outside.
Sound collection device. The sound collection device according to claim 1,
The signal processing unit
A filter estimation unit that estimates a filter that suppresses the external noise sound from the collected sound signal of the first microphone, using the collected sound signal of the first microphone and the collected sound signal of the second microphone;
A second filtering unit that performs filtering on the collected sound signal of the first microphone using the filter;
Sound collection device.

Images