JP2011179888A - Method and device for calculating wave source position - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method of calculating a wave source position, capable of enhancing calculation accuracy on a wave source position, by removing the effect of disturbance. <P>SOLUTION: In the method of calculating the wave source position, a wave signal emitted by a wave source is detected at an observation point to calculate the position of the wave source. The method includes: a step of converting the wave signal into a discrete signal of a temporal frequency; a positional estimate calculation step of calculating a positional estimate that denotes the position of the wave source for each of temporal frequencies by using the discrete signal by an expression including at least one parameter among an amplitude component of the wave signal, a differential component of the amplitude of the wave signal, and a spatial grade of the wave signal; a masking function setting step of setting a masking function that limits the band of a temporal frequency according to the value of an evaluation function by applying the positional estimate to the evaluation function denoted by using the expression; a masking step of limiting the band of a temporal frequency by the masking function; and a wave source position calculation step of calculating the position of the wave source from the positional estimate. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a wave source position calculation method and a wave source position calculation apparatus.

  A three-dimensional sound source localization sensor system based on the spatiotemporal gradient method is used to observe sound pressure, sound pressure differential and spatial sound pressure gradient at an arbitrary time at an observation point, There is known a method for estimating the position of (Non-Patent Document).

Ando et al., “Three-dimensional sound source localization sensor system based on spatiotemporal gradient method,” Proceedings of Society of Instrument and Control Engineers Vol. 29No. 5 520/528

  However, there is a possibility that the information on the sound source included in the specific time frequency section is affected by the disturbance and the estimation accuracy of the position of the sound source is lowered.

  The problem to be solved by the present invention is to provide a wave source position calculation method and apparatus capable of removing the influence of disturbance and improving the calculation accuracy of the wave source position.

  The present invention solves the above problem by calculating the position of the wave source by limiting the time frequency band of the sound signal using a masking function.

  According to the present invention, the band affected by the disturbance is identified by the masking function, the identified band is removed from the time frequency band, and the wave source position is calculated, so that the S / N ratio is improved, and the wave source The position calculation accuracy can be improved.

It is a block diagram of the wave source position calculating apparatus using the wave source position estimation method according to the embodiment of the invention. It is a calculation result of the sound source position calculation part of FIG. 1, Comprising: It is a graph which shows the frequency with respect to the azimuth angle of a detection target. It is a flowchart which shows the control procedure of the wave source position calculating apparatus of FIG. It is a graph which shows the characteristic of the threshold value ((gamma) th) of Formula (23) used with the wave source position calculating apparatus of FIG. It is a block diagram of the wave source position calculating apparatus using the wave source position estimation method according to another embodiment of the invention. It is a flowchart which shows the control procedure of the wave source position calculating apparatus of FIG. It is a figure which shows the experimental result of the azimuth | direction calculation in an Example and a comparative example.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

<< First Embodiment >>
FIG. 1 shows a block diagram of a wave source position calculation apparatus using the wave source position estimation method of this example. In this example, a vehicle is described as an example of a moving body serving as an observation point. However, this example may be applied to an apparatus other than a vehicle, and is not necessarily a moving body.

  The wave source position calculation device of this example includes a calculation device 1, a sound sensor 2, and a speed sensor 3. The arithmetic device 1 includes an integrated circuit such as a CPU and MPU, a recording medium such as an HDD, a main memory, and a cache memory. The sound sensor 2 is a sensor that detects sound emitted from another vehicle that is a sound source. The sound sensor 2 is provided with at least one general microphone, for example. The speed sensor 3 is a sensor that detects the speed of the vehicle including the wave source position calculation device of this example. The sound sensor 2 and the speed sensor 3 and the arithmetic device 1 are connected by, for example, a CAN (Controller Area Network) communication network that transmits in-vehicle signals.

  Next, the detailed configuration of the arithmetic device 1 and the control contents of the wave source position arithmetic device of this example will be described. In this example, based on the signal detected by the sound sensor 2, for example, the position estimation algorithm of Non-Patent Document 1 is applied to calculate the position of another vehicle that is a sound source.

  First, the sound sensor 2 detects a sound wave emitted from another vehicle serving as a sound source, and transmits the sound wave to the conversion unit 100 as a wave signal. In addition, since the sound sensor 2 detects the traveling sound of the host vehicle and the external sound caused by an object other than the vehicle, the wave signal includes noise from a sound source other than the detection target, and the noise is In the calculation of this example, it becomes a disturbance.

  The conversion unit 100 converts the received sound signal into a time frequency band or the like using a method such as direct conversion. Specifically, a Hann window is applied to the sound signal waveform of the received signal, the frame is divided into intervals of a constant interval every short time, and a frequency is used for each frame using a fast discrete Fourier transform (FFT). The signal is converted into a region signal S (t, ω). Here, t represents time and ω represents frequency. Thereby, the conversion unit 100 converts the wave signal into a discrete signal having a time frequency.

  The masking unit 110 removes a region that includes a disturbance and is not suitable for calculation from the signal S (t, ω) that is a discrete signal. Specifically, a masking function m (ω) set by a masking function setting unit, which will be described later, is multiplied by the signal S (t, ω) to limit the time frequency band of the signal S (t, ω). Call.

  The position estimation calculation unit 200 calculates instantaneous sound pressure, sound pressure time differentiation and sound pressure spatial gradient from the signal from the masking unit 110, and calculates a position estimation value ES (t, ω) representing the position of the sound source. To do.

  Here, the calculation method of the position estimated value ES (t, ω) by the conversion unit 100 and the position estimated value calculation unit 200 will be described in detail using the azimuth estimation algorithm of Non-Patent Document 1. The calculation target is a time frequency band that is not limited by the masking function m (ω).

  Like the spatio-temporal gradient method shown in Non-Patent Document 1, an expression that physically represents the wavefront at the observation point by the amplitude component applied to the wavefront coming from the sound source to the observation point, the differential of the amplitude component, and the spatial differential component is used. Thus, the direction and position of the wave source are estimated from the wavefront at the observation point. In order to estimate the direction of the sound source from the observation point, it is only necessary to investigate from which direction the sound has come based on the gradient.

と表現される。ｎ_ｘ、ｎ_ｙ、ｎ_ｚは、観測点から音源方向へのベクトルｎの要素であり、Ｒは観測点から音源までの距離、Ｃは音速を示す。式１乃至式３は音源が点音源であり、かつ、単一のものであれば、音圧と音圧の空間的な勾配は従属関係にある。以下では、式１乃至式３を波源拘束偏微分方程式と呼ぶ。すなわち、観測点の任意の時間における音圧、音圧時間微分および空間的な音圧勾配を計測し、波源拘束偏微分方程式を適用することで、音源の位置及び方位を演算することができる。 The space-time gradient method, when observing waves emitted from the point sound source at any point, any time of the sound pressure of the point f (t), between the sound pressure time differential When f _{t (t)} , Spatial gradients f _x (t), f _y (t), f _z (t) of the sound pressure at the observation point are

It is expressed. _nx , _ny , and _nz are elements of a vector n from the observation point to the sound source direction, R is the distance from the observation point to the sound source, and C is the speed of sound. In Expressions 1 to 3, if the sound source is a point sound source and is a single sound source, the spatial gradient of the sound pressure and the sound pressure is dependent. Below, Formula 1 thru | or Formula 3 are called a wave source restraint partial differential equation. That is, the position and orientation of the sound source can be calculated by measuring the sound pressure, the sound pressure time differential and the spatial sound pressure gradient at an arbitrary time at the observation point and applying the wave source-constant partial differential equation.

  Further, the input signal in the time domain is cut out by a window function having an appropriate time interval by the conversion unit 100 and divided into frames, and the signal converted into the frequency domain by a short-time fast Fourier transform is converted into a wave source-constrained partial differential equation. It can also be used to estimate the orientation and position of the sound source.

として表される。 From the signal S (t, ω), if the sound pressure at an arbitrary frequency is F (ω) and the sound pressure time derivative is F _t (ω), the spatial gradient of the sound pressure at the observation point is spatial. The gradients are F _x (ω), F _y (ω), F _z (ω),

Represented as:

の二乗誤差が最小となるようなｎ_ｘおよびＲを求めることを考える。その結果は、

としたとき、

で与えられる。これにより、ベクトルｎの要素であるｎ_ｘと観測点から音源までの距離Ｒが演算されるため、音源の位置が推定される。そして、上記により演算される音源の位置を示す演算値を位置推定値ＥＳ（ｔ、ω）とする。 Then, the vector n can be obtained by applying the least square method independently for each direction. However, if the horizontal plane orientation of the approaching vehicle sound source is to be obtained, n _x , n _y , which are elements of the vector n, Any one of _nz may be obtained. Focusing on a solution of n _x as an example. (Equation 4)

Consider obtaining _nx and R such that the square error of is minimized. The result is

When

Given in. Accordingly, the distance R from the n _x and the observation point are elements of the vector n to the sound source is calculated, the position of the sound source is estimated. The calculated value indicating the position of the sound source calculated as described above is assumed to be a position estimated value ES (t, ω).

  Since (Equation 8) to (Equation 13) calculate the sum for a predetermined frequency band obtained by the short-time fast Fourier transform, only the position estimation result is calculated for an arbitrary time interval. However, by calculating (Equation 8) to (Equation 13) for each frequency instead of the sum, the results of (Equation 14) to (Equation 15) can also obtain azimuth estimation results for each frequency.

により表される。また、ω_ｃは、速度センサ３の検出速度に応じて設定される。 Next, the masking unit 110, the masking function setting unit 120, and the initial setting unit 130 will be described. The initial value setting unit 130 sets the initial value of the masking function m (ω) based on the signal from the speed sensor 3. One factor of disturbance is noise generated when the host vehicle travels, and the noise affects a frictional sound with the road surface of the host vehicle, a sound generated from the host vehicle, and the like. Therefore, noise detected by the sound sensor 2 is detected according to the traveling speed of the host vehicle, and a band filter is set in advance. The bandpass filter, m (ω) [0] = 1 and then in the passband omega _c, determines at non passband m (ω) [0] = 0 and becomes as m (ω) [0],

It is represented by Further, ω _c is set according to the detection speed of the speed sensor 3.

  The masking function setting unit 120 sets the function of the band filter represented by (Equation 16) set by the initial value setting unit 130 as the masking function m (ω) and transmits the masking function to the masking unit 110.

  The masking unit 110 multiplies the masking function set by the masking function setting unit 120 with the signal S (t, ω) from the conversion unit 100 to limit the time frequency band of the signal. Thereby, the band including noise generated according to the speed of the vehicle can be excluded from the frequency band of the signal S (t, ω).

  The signal S (t, ω) filtered by the initial value of the masking function is input to the position estimation calculation unit 200, and the position estimation value ES (t, ω) is calculated as described above. The masking function setting unit 120 updates the masking function m (ω) based on the position estimated value ES (t, ω) input by the masking function setting unit 120 and outputs the updated masking function m (ω) to the masking unit 110. Thereby, this example performs feedback calculation control and improves calculation accuracy.

Hereinafter, the feedback calculation control method of this example will be described.

  The masking function set by the masking function setting unit is m (ω) [i], i is an integer of 0 or more, and indicates the number of repetitions. However, i = 0 indicates an initial value, and a function set by the initial value setting unit 130 is used.

となる。 In the position estimation unit calculation unit 210, by applying the masking function m (ω) [i] and expressing (Expression 14) to (Expression 14),

It becomes.

Then, (Equation 17) to (Expression 22), using (Equation 14) and (Equation 15), and calculates the distance R from the _{n x} and the observation point is the x direction component of the vector n to the sound source, The position estimated value ES (t, ω) is calculated. The estimated position value ES (t, ω) is input to the masking function setting unit 120. The masking function setting unit 120 uses the error value ER (t, ω) between the theoretical value and the actually calculated value as the wave source. Calculation is performed using a constrained partial differential equation.

のように表され、位置推定値ＥＳ（ｔ、ω）を（式２３）に代入する。 The error value ER (t, ω) is expressed using a wave source-constrained partial differential equation:

The position estimated value ES (t, ω) is substituted into (Equation 23).

  Next, the masking function setting unit 120 sets the masking function m (ω) [i + 1] from the error value ER (t, ω).

のように、ER(t、ω)の閾値γ_ｔｈに対する大きさによって表される。閾値γ_ｔｈは閾値γ_ｔｈ（ω）とし、帯域毎に車速に応じた値である。例えば、車速が所定の速度より低いときには１ｋＨｚ未満のγ_ｔｈ（ω）を閾値Ｃ_Ｌに、１ｋＨｚ以上のγ_ｔｈ（ω）を閾値Ｃ_Ｈ（＞Ｃ_Ｌ）にする。車速が当該所定の速度より高いときには２ｋＨｚ未満のγ_ｔｈ（ω）を閾値Ｃ_Ｌに、２ｋＨｚ以上のγ_ｔｈ（ω）を閾値Ｃ_Ｈ（＞Ｃ_Ｌ）にする。（式２４）より、誤差値ＥＲ（ｔ、ω）は、時間周波数帯域における、理論値との誤差を示す評価関数であって、誤差値ＥＲ（ｔ、ω）が大きければ、外乱が含まれる可能性が高く、誤差値ＥＲ（ｔ、ω）が小さけば、外乱が含まれる可能性が低いと評価される。そして、（式２４）に示すとおり、マスキング関数ｍ（ω）［ｉ＋１］は、誤差値ＥＲ（ｔ、ω）が閾値γ_ｔｈより大きい帯域に制限をかけ、誤差値ＥＲ（ｔ、ω）が閾値γ_ｔｈより小さい帯域に制限をかける関数とする。なお、閾値γ_ｔｈは、音波の振幅レベルに応じて決定してもよい。 The masking function m (ω) [i + 1] is

As described above, it is represented by the magnitude of ER (t, ω) with respect to the threshold value γ _th . The threshold value γ _th is a threshold value γ _th (ω), and is a value corresponding to the vehicle speed for each band. For example, when the vehicle speed is lower than a predetermined speed, γ _th (ω) of less than 1 kHz is set to the threshold value C _L and γ _th (ω) of 1 kHz or more is set to the threshold value C _H (> C _L ). The vehicle speed is less than 2kHz gamma _th to (omega) the threshold _{C L} when higher than the predetermined speed, to 2kHz or more gamma _th (omega) the threshold _C H (> _{C L).} From (Equation 24), the error value ER (t, ω) is an evaluation function indicating an error from the theoretical value in the time-frequency band. If the error value ER (t, ω) is large, disturbance is included. If the possibility is high and the error value ER (t, ω) is small, it is evaluated that the possibility that a disturbance is included is low. Then, as shown in (Equation 24), the masking function m (ω) [i + 1 ] , the error value ER (t, omega) Place limits the threshold gamma _th larger bandwidth, error value ER (t, omega) is A function that limits a band smaller than the threshold γ _th is used. The threshold γ _th may be determined according to the amplitude level of the sound wave.

  Then, the masking function setting unit 120 updates the masking function m (ω) [i] to the masking function m (ω) [i + 1] based on (Equation 24), and outputs it to the masking unit 110. The masking unit 110 uses the masking function m (ω) [i + 1] to limit the time frequency band of the signal from the conversion unit 100, and the position estimated value calculation unit 210 performs the position estimated value ES (t, ω ) Is calculated.

  The masking function setting unit 120 also determines whether the masking function m (ω) [i + 1] has converged. The masking function setting unit 120 stores the previous masking function m (ω) [i], and the masking function has converged when m (ω) [i + 1] = m (ω) [i]. to decide. Note that the convergence determination may be made, for example, when the preset maximum value max (i) of i and the number of repetitions i are equal to each other.

  When the masking function setting unit 120 determines that the masking function has converged, the control of this example is skipped from the feedback calculation control described above, and the following clustering process by the sound source position calculation unit 310 is performed.

  The sound source position calculation unit 310 statistically calculates a position estimated value for each frequency from the position estimated value ES (t, ω), groups all azimuths to be detected for each arbitrary width, This region is converted into a histogram using bins, and the position of the sound source is calculated. An example of the calculation result is shown in FIG. FIG. 2 is a graph showing the frequency with respect to the azimuth angle of the detection target. In this example, the azimuth to be detected is set between −90 ° and + 90 °, and is grouped into six regions. Then, based on the position estimation value ES (t, ω), the azimuth is calculated for each frame within a certain time, assigned to the six regions, and the assigned frequency is counted. The frequency in FIG. 2 corresponds to the counted frequency. Then, the sound source position calculation unit 310 calculates the position of the sound source from the position estimation value in the frequently used region. For example, in FIG. 2, the sound source position is calculated by calculating the distance R from the position estimation value belonging to the azimuth of 30 to 60 ° with the high frequency of 30 to 60 ° as the azimuth.

  Next, the control procedure of this example will be described with reference to FIG. FIG. 3 is a flowchart showing the control procedure of the wave source position calculation apparatus of this example, and shows the wave source position calculation method of this example.

  First, in step S1, the sound sensor 2 detects sound waves in the environment outside the vehicle, and the speed sensor 3 detects the speed of the host vehicle. In step S2, the conversion unit 100 converts the wave signal of the sound wave detected in step S1 into a discrete signal S (t, ω). Next, in step S3, the initial value setting unit 130 sets the initial value m (ω) [0] of the masking function based on the vehicle speed detected in step S1.

  In step S4, the masking unit 110 limits the time frequency band S (t, ω) of the discrete signal using the set masking function m (ω) [0]. In step S5, the position estimated value calculation unit 200 calculates a position estimated value ES (t, ω) [i] from the limited discrete signal. Here, i represents the number of repetitions from step S4 to step 8, and the number of repetitions immediately after setting the initial value in step S3 is set to i = 0.

  Next, in step S6, the masking function setting unit 120 compares the theoretical value recorded in advance on the recording medium or the like with the position estimated value calculated in step S5, and calculates an error value ER ( t, ω) [i] is calculated. In step S7, the masking function setting unit 120 sets the masking function m (ω) [i + 1] based on the calculated error value ER (t, ω). In step S8, the masking setting unit 120 compares the masking function m (ω) [i + 1] with the masking function m (ω) [i] to determine whether the masking function has converged.

  If the masking function has not converged, the process returns to step S4, and the masking unit 110 uses the masking function m (ω) [i + 1] set in step S7 to use the time frequency band S (t, ω of the discrete signal. In step S5, the position estimated value calculation unit 200 calculates a position estimated value ES (t, ω) [i + 1] from the limited discrete signal.

  On the other hand, when the masking function has converged, the sound source position calculation unit 310 calculates the position of the sound source from the position estimated value ES (t, ω) [i] calculated in step S7, and ends the processing of this example. To do.

  As described above, the wave source position calculation method or the wave source position calculation apparatus of this example uses discrete signals, and at least the amplitude component of the wave signal, the differential component of the amplitude of the wave signal, or the spatial gradient of the wave signal. A position estimation value indicating the position of the sound source is calculated by an expression including one parameter. Then, the position estimation value is substituted into the evaluation function expressed using the equation, a masking function is set according to the value of the evaluation function, and the time frequency band of the wave signal is limited by the masking function. Thus, in this example, the frequency band including the disturbance is efficiently identified, and the time frequency band is limited in the process of calculating the position of the sound source, so the S / N ratio is improved and the position of the sound source is calculated. Accuracy can be increased. Also, in the process of calculating the position of the sound source, the band including the disturbance is removed, and the calculation in the band that lowers the accuracy of the sound source position is not performed, so the accuracy of calculating the position of the sound source is reduced while reducing the calculation load. Can be increased.

  This example also compares the ideal value of the estimated position value with the estimated position value that is actually measured, uses an evaluation function to calculate the error value, identifies the band that includes the disturbance, applies feedback, The frequency band is limited and the position of the sound source is calculated. As a result, the frequency band including disturbance is efficiently identified, and the time frequency band is limited in the process of calculating the position of the sound source, so the S / N ratio is improved and the accuracy of calculating the position of the sound source is increased. Can do.

  Moreover, this example evaluates the value of the said evaluation function by comparing with the threshold value based on the speed of the own vehicle. As a result, the frequency band including the disturbance can be efficiently identified in accordance with the change in the speed of the host vehicle, so that the accuracy of calculating the position of the sound source can be increased.

  In this example, the speed of the host vehicle is reflected in the initial value of the masking function. As a result, the frequency band including the disturbance can be efficiently identified in accordance with the change in the speed of the host vehicle, so that the accuracy of calculating the position of the sound source can be increased.

  In this example, the sound pressure from the wave signal of the sound source, the sound pressure time derivative, and the spatial gradient of the sound pressure are used as parameters, and the wave source restraint partial differential equation is represented. Instead of the spatial gradient, particle velocity, medium impedance, or the like may be used.

となる。この現象を粒子速度で表現すると、

と表現できる。ここで、ρ_０は平均密度、ｃは音速であり、ｆとｇは、空間的及び時間的境界条件に依存する関数であって、それぞれｘの正及び負方向への速度ｃで伝播する波を示す。更に、音圧と粒子速度の間には、媒質の音響インピーダンスを用いて、

の関係がある。すなわち、粒子速度及び媒質の音響インピーダンスが分かれば、音圧を演算することができる（例えば文献 Ｐ．Ａ．Ｎｅｌｓｏｎ＆Ｓ．Ｊ．Ｅｌｌｉｏｔｔ、“Ａｃｔｉｖｅ Ｃｏｎｔｒｏｌ ｏｆ Ｓｏｕｎｄ”、ｐ１６、ｐ１７を参照（文献３））。また、（式２５）、（式２６）及び（式２７）は、３次元に拡張できる（文献２及び３を参照）。そのため、音圧、音圧時間微分及び音圧の空間勾配は、粒子速度や音響インピーダンスを用いて演算することができ、（式１）〜（式３）を導出することができる。なお、粒子速度は、周知の方法（例えば、ｈｔｔｐ：／／ｗｗｗ．ｍｉｃｒｏｆｌｏｗｎ．ｃｏｍ／ または ｈｔｔｐ：／／ｔｏｙｏ．ｃｏ．ｊｐ／pｃｂ／ｏｎｋｙｏｒｕｓｈｉ／ 要約すると、空気分子の振動速度を直接計測する音響粒子速度センサ）を用いることにより検出される。 For example, using sound pressure, the sound field (that is, one-dimensional wave) inside a long tube with a rigid and uniform wall is represented by sound pressure (eg, Hideki Tachibana, “Sound Intensity” Ohm p26) -37 (Reference 2)), when the sound pressures of the traveling wave and the backward wave are expressed by functions of f and g, respectively, the sound pressure at an arbitrary point is represented by a general solution of a one-dimensional wave equation,

It becomes. Expressing this phenomenon in terms of particle velocity,

Can be expressed as Here, ρ ₀ is the average density, c is the speed of sound, and f and g are functions that depend on spatial and temporal boundary conditions, and waves that propagate at a velocity c in the positive and negative directions of x, respectively. Indicates. Furthermore, between the sound pressure and the particle velocity, using the acoustic impedance of the medium,

There is a relationship. That is, if the particle velocity and the acoustic impedance of the medium are known, the sound pressure can be calculated (see, for example, the document PA Nelson & SJ Elliott, “Active Control of Sound”, p16, p17 (reference 3). ). Also, (Equation 25), (Equation 26), and (Equation 27) can be expanded to three dimensions (see Documents 2 and 3). Therefore, the sound pressure, the sound pressure time derivative, and the spatial gradient of the sound pressure can be calculated using the particle velocity and the acoustic impedance, and (Expression 1) to (Expression 3) can be derived. It should be noted that the particle velocity is measured by a well-known method (for example, http://www.microflow.com/ or http://toyo.co.jp/pcb/onkyorusi/ It is detected by using a particle velocity sensor.

  Further, in this example, the sound pressure time derivative may be directly observed by a well-known method (for example, the documents N. Ono, A. Saito, and S. Ando, “Bio-mound Sound Source Localization Diaphragm” IE Trans.SM, Bol.123 No. 3, 2003).

  The wave-constant partial differential equation of this example is expressed by other equations, for example, using a sound field expression method using sound pressure (wave amplitude) or particle velocity developed from the wave equation, or a combination thereof. And expressed.

  In this example, the wave source constrained partial differential equation is expressed using the amplitude component of the wave signal, the differential component of the amplitude of the wave signal, and the spatial gradient parameter of the wave signal. The wave source light source differential equation may be expressed by at least one parameter among the differential component of the amplitude of the wave signal and the spatial gradient parameter of the wave signal.

  In this example, instead of the above-mentioned wave source-constrained partial differential equation, the sound source and the observation point only have to be expressed as an equation having a desired relationship. For example, in a region where a tunnel such as an automobile or an outer wall of a highway exists An equation based on a virtual image method using a duct as a model may be used (see Document 2).

  In this example, the position estimated value calculation unit 210 or the sound source position calculation unit 310 calculates the distance from the observation point to the sound source based on the signal of the sound sensor 2, but is measured using a camera, a radio transceiver, or the like. May be.

  In this example, the initial value setting unit 130 sets the initial value of the masking function m (ω) according to the detection speed from the speed sensor 3. The initial value is set as a constant value set in advance. The initial value of the masking function m (ω) may be set so that the signal is allowed to pass in a predetermined frequency band and the signal is limited in other frequency bands. In this example, the initial value of the masking function m (ω) may be set according to the parameter calculated by the position estimation value calculation unit. In this example, the speed sensor 3 may be omitted.

としてｍ(ω)を設定してもよい。なおＳ_ｔｈは、予め設定される閾値である、なお、離散信号Ｓ（ｔ、ω）の大きさは、離散信号の波面の瞬時振幅を用いてもよいし、当該瞬時振幅の微分値を用いてもよい。 In this example, the masking function setting unit 120 and the initial setting unit 130 preset the passband ω _c as the initial value of the masking function and update the masking function, but the discrete signal S (t, ω) for each band is updated. Depending on the magnitude of (for example, the absolute value of the amplitude)

M (ω) may be set as Note that S _th is a preset threshold value. Note that the magnitude of the discrete signal S (t, ω) may be the instantaneous amplitude of the wave front of the discrete signal, or a differential value of the instantaneous amplitude. May be.

ここでεは任意の微小な値である。これにより、（式２３）の発散を防ぐことができる。 In Equation 23, when the value of the denominator | F _x (τ, ω) | ² approaches 0 (that is, a wave source coming from a direction perpendicular to the x axis), the error ER (t, ω) is ideal. It diverges even in the state. Therefore, when the value of the denominator | F _x (τ, ω) | ² approaches 0,
The following formula is introduced.

Here, ε is an arbitrary minute value. Thereby, the divergence of (Formula 23) can be prevented.

As countermeasures against the wave source arriving from a direction orthogonal to the x axis, it may be a function which depends the γth equation (24) into n _x. FIG. 4 is a graph showing the characteristic of the threshold value (γth) with respect to the x-axis angle in the calculation result of the calculation unit 310. For example, when the direction orthogonal to the x-axis is set to 0 °, the value of γth is set to be larger than the value of the angle region other than near 0 ° as shown in FIG. Then, when the calculation result of (Expression 23) diverges, the masking function setting unit 120 sets the characteristic shown in FIG. 4 to the threshold value (γth) of (Expression 23), and calculates the masking function m (ω). Thereby, the position of the sound source in the direction orthogonal to the x-axis can be calculated without using (Equation 29).

  In (Expression 23) of this example, the square of the absolute value of the spatial gradient of the sound pressure is set in the denominator in order to normalize the wave source constrained partial differential equation, but the denominator is not necessarily required. The masking function is a function having at least one of an error value ER (t, ω), an inverse number of the error value ER (t, ω), or a value obtained by normalizing the error value ER (t, ω) as an element. I just need it.

  In this example, the observation point is a sound source, and the sound sensor 2 detects the sound wave. However, the observation target is not limited to the sound source, and may be a wave source. The sound sensor 2 may detect a wave from the wave source. Good. The wave source position calculation device and the wave source position calculation method of the present example described above can be applied not only to a sound source but also to a wave source to calculate the position of the wave source.

  The conversion unit 100 of this example corresponds to the “conversion unit” of the present invention, the position estimation value calculation unit 200 is the “position estimation value calculation unit”, the masking function setting unit 120 is the “masking function setting unit”, The masking section 110 corresponds to “masking means”, and the sound source position calculation section 310 corresponds to “wave source position calculation means”.

<< Second Embodiment >>
A wave source position calculation apparatus and a wave source position calculation method according to another embodiment of the invention will be described with reference to FIG. This example differs from the first embodiment described above in part of the control content in the arithmetic device 1. The description of the same configuration as that of the first embodiment described above in other configurations is incorporated as appropriate. FIG. 5 shows a block diagram of the wave source position calculation apparatus of the present invention.

  As shown in FIG. 5, the basic configuration is the same as that of the wave source position calculation apparatus of the first embodiment except that the speed sensor 3 and the initial value setting unit 130 are not provided, but the control procedure is different.

  The masking function setting unit 120 uses the (Equation 17) to (Equation 23) based on the position estimated value ES (t, ω) calculated by the position estimated value calculating unit 200 to use the error value ER (t, ω ) And a masking function m (ω) is calculated based on (Equation 24).

  The masking unit 110 uses the masking function m (ω) to limit the frequency band of the position estimated value ES (t, ω) input from the position estimated value calculating unit 200. Then, the sound source position calculation unit 310 statistically determines the azimuth angle of the sound source and the observation from the position estimation value ES (t, ω) of the band that has passed the masking function m (ω), as in the first embodiment described above. The distance from the point to the sound source is calculated, and the position of the sound source is calculated.

  Next, the control procedure of this example will be described with reference to FIG. FIG. 6 shows a control procedure of the wave source position calculation apparatus of this example, and shows a flowchart of the wave source position calculation method of this example.

  First, in step S11, the sound sensor 2 detects sound waves in the environment outside the vehicle. In step S12, the conversion unit 100 converts the wave signal of the sound wave detected in step S11 into a discrete signal S (t, ω). In step S13, position estimated value calculation section 200 calculates position estimated value ES (t, ω) from the discrete signal. Next, in step S14, the masking function setting unit 120 compares the theoretical value recorded in advance on the recording medium or the like with the position estimated value calculated in step S13, and calculates an error value ER ( t, ω) is calculated.

  In step S15, the masking function setting unit 120 sets the masking function m (ω) based on the calculated error value ER (t, ω). In step S <b> 16, the masking unit 110 uses the set masking function m (ω) to limit the time frequency band ES (t, ω) of the position estimation value. In step S17, the position of the sound source is calculated from the limited position estimation value ES (t, ω), and the process of this example ends.

  As described above, this example uses the parameter included in the equation having at least one of the differential component of the amplitude of the wave signal or the spatial gradient of the wave signal as a parameter. ), The band including the disturbance is specified, the position of the sound source is calculated by limiting the frequency band of the position estimation value ES (t, ω).

  Thus, in this example, the frequency band including the disturbance is efficiently identified, and the time frequency band is limited in the process of calculating the position of the sound source, so the S / N ratio is improved and the position of the sound source is calculated. Accuracy can be increased. Also, in the process of calculating the position of the sound source, the band including the disturbance is removed, and the calculation in the band that lowers the accuracy of the sound source position is not performed, so the accuracy of calculating the position of the sound source is reduced while reducing the calculation load. Can be increased.

In this example, the masking function setting unit 120 may use a function that performs weighting according to the time frequency band. For example, in the bands of 1 kHz, 1.5 kHz, and 2 kHz, the masking function m ₁ (ω) is m ₁ (1 kHz) = 2, m ₁ (1.5 kHz) = 0.8, and m ₁ (2 kHz) = 1. The new masking function is set by multiplying the masking function based on the error value ER (t, ω). Thereby, for example, when the calculation result by the calculation unit 310 is the frequency value F = 1 in the direction of 50 ° without weighting, the frequency value F = 1 × 2 + 1 × 0.8 + 1 × 1 = 3 in this example. .8 can be weighted.

<< Third Embodiment >>
A wave source position calculation apparatus and a wave source position calculation method according to another embodiment of the invention will be described. This example differs from the first or second embodiment described above in the masking function setting unit 120 in the method of calculating an error value that is an evaluation function and the masking function.

The description of the same configuration as that of the first or second embodiment described above in other configurations is incorporated as appropriate.

のように、複数の時間フレームを示すτを導入することで時間周波数領域に対して最適な帯域を通過させるフィルタを設計することができる。 In the masking function setting unit 120 of this example, the masking function m (ω) can also reflect statistics over a plurality of short-time frames. For example,

Thus, by introducing τ indicating a plurality of time frames, it is possible to design a filter that passes an optimum band with respect to the time frequency domain.

  Thus, in this example, since a plurality of time frames can be statistically reflected in the masking function m (ω), the frequency band including the disturbance can be efficiently identified, and the S / N ratio is improved. The accuracy of calculating the position of the sound source can be increased.

  Examples that further embody the present invention will be described below.

  The wave source position calculation device of this example was mounted on a vehicle, traveled on a test road, calculated the sound source position of another vehicle, and evaluated the calculation results. The host vehicle was provided with two microphones. Two other vehicles to be measured are prepared. The first vehicle overtakes the host vehicle from behind, and the second vehicle approaches from the front of the host vehicle and travels behind the host vehicle. Then, the masking function m (ω) is calculated using (Expression 30) to (Expression 35), (Expression 37), and (Expression 29), and the sound wave discrete signal is calculated using the masking function m (ω). The position estimated value ES (t, ω) was calculated again, and the calculated azimuth was plotted.

  In the comparative example, the position of the microphone mounted on the vehicle, the other vehicle to be measured, and the traveling conditions of the other vehicle are the same, but the calculation based on the masking function m (ω) and the signal based on the masking function m (ω) The time frequency band was not filtered, the position estimation value was calculated from the signal without band limitation of this example using the light source constrained partial differential equation, and the calculated direction was plotted.

  In the example and the comparative example, the recording conditions are a sampling frequency of 48 kHz and a quantization bit number of 16 bits. The recorded acoustic signal does not have much fluid noise of concern. In addition to the above, it is confirmed that the driving sound of other vehicles can be acquired. The analysis conditions were such that the frame length was 128 points (2.67 ms) and the frame shift was 64 points (133 ms) in order to obtain a high time resolution.

The plotted results are shown in FIG. FIG. 7 shows the orientation of the measurement sound source with respect to time. In addition, the azimuth | direction of a vertical axis | shaft shall be the value (Nx) which normalized the angle to -1.0- + 1.0. The horizontal axis indicates time. In FIG. 7, ◯ indicates the value of the example, and X indicates the value of the comparative example. In addition, the Δ mark and the □ mark indicate that the position coordinates of the vehicle are measured by GPS and converted to the direction of another vehicle based on the position of the own vehicle.
≪Discussion≫

  As shown in FIG. 7, the detection range of the example indicated by the arrow A is wider than the detection range of the comparative example indicated by the arrow B. As a result, since the embodiment calculates the direction of the sound source by removing the area determined to contain a lot of noise, the S / N ratio is improved, the angular range for following the sound source is widened, and the distance of other vehicles is increased. It can be confirmed that correct calculation is obtained up to a far range.

  Based on the above results, the masking function is calculated to specify the band including the disturbance, and the position of the sound source is calculated based on the signal excluding the specified band, thereby improving the detection accuracy and the detection target range. Can be spread.

DESCRIPTION OF SYMBOLS 1 ... Arithmetic apparatus 2 ... Sound sensor 3 ... Speed sensor 100 ... Conversion part 110 ... Masking part 120 ... Masking function setting part 130 ... Initial value setting part 200 ... Position estimation value calculation part 310 ... Calculation part

Claims (10)

A wave source position calculation method for detecting a wave signal emitted by a wave source at an observation point and calculating a position of the wave source,
Converting the wave signal into a time-frequency discrete signal;
Using the discrete signal, the wave signal is generated for each time frequency according to an expression including at least one parameter of an amplitude component of the wave signal, a differential component of the amplitude of the wave signal, or a spatial gradient of the wave signal. A position estimated value calculating step for calculating a position estimated value representing the position of the source;
A masking function setting step for setting a masking function for limiting the band of the time frequency according to the value of the evaluation function by substituting the position estimation value into the evaluation function represented using the equation;
A masking step of limiting the band of the time frequency by the masking function;
And a wave source position calculating step of calculating a position of the wave source from the estimated position value. 2. The wave source position calculating method according to claim 1, wherein the position estimated value calculating step calculates the position estimated value using the discrete signal in which a time frequency band is limited by the masking step. . The masking step places a limit on the time frequency band of the position estimated value calculated by the position estimated value calculating step,
3. The wave source position according to claim 1, wherein the wave source position calculating step calculates the position of the wave source from the position estimation value in which a time frequency band is limited by the masking step. Calculation method. The value of the evaluation function includes a position estimated value (ES [i] (where i is an integer equal to or greater than 0)) calculated by the position estimated value calculating step, and an ideal value calculated in advance using the equation. Error value indicating the error of
The masking function setting step sets a masking function (m [i + 1]) based on the error value,
The position estimated value calculating step calculates a position estimated value (ES [i + 1]) using the discrete signal with a limited time frequency band based on the masking function (m [i + 1]). The wave source position calculation method according to claim 1, wherein: 5. The wave source position calculating method according to claim 1, wherein the masking function setting step compares the value of the evaluation function with a predetermined value and sets the masking function. 6. . The masking function (m [i]) is the error value (ER [i]), the reciprocal of the error value (ER [i]), or a value obtained by normalizing the error value (ER [i]). 5. The wave source position calculation method according to claim 4, wherein the function is a function having at least one element. The initial value of the masking function is
A value set according to the amplitude component of the wave signal, the differential component of the amplitude of the wave signal,
The wave source position calculation method according to any one of claims 1 to 6, wherein the wave source position calculation method is a value set according to a moving speed of the observation point or a preset value. Detecting a speed of a moving object including the observation point;
Setting a noise band including a running noise of the moving body according to the detection speed;
The wave source position calculation method according to claim 1, further comprising: removing the noise band from the time frequency band. Detecting the speed of the moving object including the observation point,
6. The wave source position calculation method according to claim 5, wherein the masking function setting step sets the predetermined value according to the detection speed. A wave source position calculation device that detects a wave signal emitted by a wave source at an observation point and calculates the position of the wave source,
Conversion means for converting the wave signal into a discrete signal of time frequency;
Using the discrete signal, the wave signal is generated for each time frequency according to an expression including at least one parameter of an amplitude component of the wave signal, a differential component of the amplitude of the wave signal, or a spatial gradient of the wave signal. Position estimated value calculating means for calculating a position estimated value representing the position of the source;
Masking function setting means for setting a masking function for limiting the band of the time frequency according to the value of the evaluation function by substituting the position estimation value into the evaluation function represented by the equation;
Masking means for limiting the band of the time frequency by the masking function;
And a wave source position calculating means for calculating a position of the wave source from the position estimated value.

Images