JPH07146350A - Method for orienting azimuth of sound source - Google Patents

Abstract

PURPOSE:To orient the azimuth of a sound source with high accuracy by receiving acoustic wave signals by means of multiple receiving elements and spectrum- estimation of the spatial frequency generated by the sequence of numbers of the output signal of each receiving element by the maximum entropy method (MEM). CONSTITUTION:Input signals l, 2, 3{X1(t), X,(t),..., Xm(t)} from linearly arranged receiving elements are converted into digital signals by means of A/D converters 11, 12, and 13. Then the digital signal trains (sequences of numbers formed in accordance with the order of the elements) at the same time ni:DELTAt, namely, sequences of numbers {Xj(ni:DELTAt)}m composed of (m) pieces of data are stored in a memory 14 composed of a RAM, etc. A computing element 15 which performs spectrum-presumption from the sequences of numbers by using the MEM finds a power spectrum composed of pieces of power. A detector 16 which detects the position (LQSP) of a spectral peak from the power spectrum decides the spatial frequency fp having the peak of power. Then an azimuth converter 17 orient the azimuth of a sound source from the spatial frequency fp.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of estimating the direction of a sound source from a signal from a single sound source by spectral analysis of signals from a plurality of wave receiving elements at the same time, and more particularly to a method of receiving a plurality of received waves arranged in a straight line. The present invention relates to a method for estimating a sound source direction with high accuracy by spectral estimation using a maximum entropy method for elements.

[0002]

2. Description of the Related Art When a sound source direction is estimated by spectrum analysis, a sound source at a distant position arrives in a linear array as a plane wave as shown in FIG. 2, and therefore, as schematically shown in FIG. The sequence of output signals is
Has a spatial frequency. Between the sound source direction θ and the spatial frequency f (cgcle / cm) of the sequence, θ = sin ⁻¹ (f / f
90) expresses the relationship. Here, f90 is the spatial frequency when the wavefront of the sound source is perpendicular to the linear array. Conventionally, the spatial frequency has been estimated from the peak value of the power spectrum by Fourier transform (or fast Fourier transform) of the received signal sequence.

[0003]

The above-described conventional method for estimating the sound source direction by obtaining the spectrum by using the Fourier analysis requires more receiving elements when obtaining the spatial frequency with higher resolution. There is a drawback that The present invention provides a method for reducing the number of receiving elements by estimating the spectrum by using the maximum entropy method, and providing a method for reducing the variation of spectral peaks by the maximum entropy method.

[0004]

Since the spectrum estimation of the present invention uses the maximum entropy method, the number of receiving elements can be small. Also, the maximum entropy method, for example, Bur
In the g method, the position of the spectral peak (Locatio
no of Spectral Peak, LO, SP)
Varies according to the phase of the received signal sequence, so a uniform random number is generated and arbitrary) time (hence arbitrary phase)
A large number of received signal trains are calculated and their averages are taken to improve the reliability of the direction estimation value. The maximum entropy method by which the spectrum can be estimated even if the number of receiving elements (the number of data) is small is specifically obtained by the following equation.

[0005]

[Equation 1]

Here, P: power spectrum, f: spatial frequency, Δt: sampling period, Pm: prediction error average,
γmk: coefficient of line prediction filter. (For example, Mikio Hino, "Spectral Analysis", Asakura Shoten (1977) pp.
210-221). The relationship between the movement of LOSP and the phase of data in the Burg method is described in W. Y Chen and
G. R. Stegen "Experiments w
it maximum power
spectra of sinusoids, "J.G.
eophys. Res. , Val: 79, pp. 301
See 9-3022 July, 1974.

[0007]

The present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of the present invention.
Input signals 1 from the linear array receiving elements shown in FIG.
2, 3 (X ₁ (t), X ₂ (t), X _m (t) are A / D converters 1
A / D conversion by 1, 12, 13 and at the same time n
i: A digital signal sequence of Δt (a sequence of numbers created in the order of elements), that is, a sequence of numbers {Xj (n
i: Δt} m is stored in a memory such as a RAM. This sequence is a calculator 1 that estimates the spectrum by the maximum entropy method.
5, a power spectrum composed of M powers {P (N · δf)} M (where N = 1, 2, ... M, δf is the frequency resolution of the power spectrum as necessary) is obtained. The spatial frequency fp having the power peak is determined by the LOSP detector 16 which detects the position of the spectral peak from the power spectrum. fp
The azimuth converter 17 is used to obtain the estimated azimuth. (In the case of the linear array, as described above, the relationship between the spatial frequency and the sound source direction is θp = sin ⁻¹ (fp / f90).)
Furthermore, the maximum entropy method has the advantage that the power spectrum can be estimated with high frequency resolution with short data,
On the other hand, LOSP is L depending on the phase of the data of the data length.
The OSP shifts from the true spatial frequency.

The deviation from the spatial frequency value of the vacuum can be reduced by increasing the data length, or by fixing the data length and averaging the LOSPs obtained by the sequence having different phases. In the present invention, the data at time ni · Δt is added K times (with reference to the inclusion of all phases) by the adder 18 and data having different phases is selected by uniform random number generation. Select by 20. When it is determined that the azimuth θp calculated through K operations has ended (19), the multiplier 21 multiplies 1 / K to output the estimated azimuth θ.

The flowchart of the maximum entropy method spectrum estimation calculator is shown in FIG. In FIG. 4, the prediction error average Pm by the Burg method and M average filter coefficients γmk
The following is an algorithm for recursively obtaining the Levinson algorithm.

[0010]

As described above, according to the present invention, in order to reduce the deviation from the true spatial frequency by estimating the sound source direction from the estimation of the spectrum of the signal of the few receiving elements by the maximum entropy method, many By taking the arithmetic mean of the estimated values from the data at different times, there is an effect that the direction can be estimated with high resolution and high accuracy.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention.

FIG. 2 is a schematic diagram showing a geometrical relationship with respect to a wavefront of a wave receiving element having a linear array used in the present embodiment, where θ is a sound source direction, λ is a wavelength of a sound wave, and d is an interval between elements. , Where d <λ / 2.

FIG. 3 schematically shows a sequence of output signals of the wave receiving element at the same time.

FIG. 4 is a flowchart showing an algorithm of the Burg method, which is one of the maximum entropy methods.

[Explanation of symbols]

1, 2 and 3 Input signals from elements 1, 2 and m 11, 12, 13 A / D conversion 14 Memory for storing m data 15 Operator for spectrum estimation by maximum entropy method 16 Spectral peak position from power spectrum Detector 17 that detects and outputs the estimated spatial frequency 17 azimuth converter 18 that calculates the azimuth from the spatial frequency 18 adder 19 Device that counts the number of calculations and determines the number of calculations 20 Generates a uniform random number and determines the time Device 21 Multiplier

Claims (1)

[Claims] 1. A maximum entropy method (Maximum) is used to obtain a spatial frequency obtained by receiving a sound wave signal from a single distant sound source by a plurality of wave receiving elements arranged in space and using a sequence of output signals of the wave receiving elements at the same time. Entropy Metho
d, MEM) to estimate the spectrum, estimate the direction of the sound source from the maximum value of the power spectrum with respect to the obtained spatial frequency, and take the average of the directions estimated from the sequence of the outputs of the receiving elements at many different times. A sound source direction estimation method characterized by reducing variations in directions.