JP2576044B2 - Signal tracking method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an unknown narrow direction having a directivity based on a signal intensity distribution in a frequency space obtained by performing a frequency analysis on the output of a receiving beam directed in a plurality of directions. The present invention relates to a signal tracking method for measuring a band signal continuously over time.

[0002]

2. Description of the Related Art A conventional method for tracking a narrow band signal will be described. A signal is received by a plurality of receiving beams pointing in directions slightly different from each other, and frequency analysis is performed on each output of the receiving beam to perform spectrum estimation at each sample time. As a result, the signal intensity X at each frequency f of the input signal of the receiving beam directed in each direction θ is obtained.
(F, θ) is obtained. Next, a peak of the signal strength X (f, θ) is detected. Here, the peak is a local maximum point at which the signal intensity is greater than the intensity at the adjacent sample frequency in the same receiving beam and the azimuth is greater than the intensity at the same sample frequency of the receiving beam having both directions. . If the point (f0, θ0) is at a peak and the signal intensity exceeds a predetermined threshold, it is determined that a narrow band signal having a frequency f0 has arrived, and the azimuth is approximately θ0. to decide. Subsequently, as a more accurate measurement of the direction of arrival of the narrowband signal, the signal intensity X
From the data near the point (f0, θ0) at (f, θ), the precise measurement direction θ1 is obtained. The precise measurement direction θ1 is, for example, the frequency f0 of a plurality of reception beams directed to a direction near the direction θ0.
Is obtained by performing an interpolation process by using the signal strength in the above to obtain the maximum point. In this manner, it is detected that the narrow band signal of the frequency f0 arrives from the precise measurement direction θ1. After performing the above processing for each sample time, the frequency and precise measurement direction of the narrowband signal detected at different times are compared. The difference between the frequency f0 and the arrival direction θ1 of the narrowband signal detected at a certain sample time and the frequency f0 ′ and the arrival direction θ1 ′ of the narrowband signal detected within a predetermined time in the past is equal to or less than a predetermined value. In some cases, both signals are considered and integrated as narrowband signals from the same source. Thereafter, by repeating in the same manner, the ever-changing frequency and arrival direction of the narrowband signal are obtained over time, and the narrowband signal is tracked.

[0003]

However, the conventional signal tracking method has the following problems. The situation where the signal strength is small compared to the noise strength, ie, S /
In a situation where the N-ratio is low, a peak based on random noise frequently appears in the distribution of the signal strength, together with a true peak based on the narrowband signal. Therefore, in the conventional signal tracking method, since a phenomenon that a peak based on noise is often integrated into a narrowband signal being tracked often occurs, a problem that an incorrect frequency or arrival direction is indicated or tracking is interrupted. Yes, it could not be used unless the signal strength was sufficiently large compared to the noise strength. 2A and 2B are diagrams showing the above problem and showing the signal intensity distribution in the azimuth space with respect to the frequency f. FIG. 2A shows the signal intensity distribution at the sample time T1, and FIG. 2B shows the signal intensity distribution at the sample time T2. FIG. 3C shows a signal intensity distribution diagram at the sample time T3, and FIG. 3D shows the sample time T4.
FIG. 4 is a signal intensity distribution diagram at. FIG. 3 is a diagram showing a desirable tracking result of the narrow band signal shown in FIG. 2, and FIG.
FIG. 9 is a diagram showing a conventional tracking result of the narrowband signal shown in FIG. As shown in FIG. 2A, at the sampling time T1, a narrow band signal arrives from the azimuth θ1. As shown in FIG. 2B, at the sample time T2, a narrowband signal arrives from the azimuth θ2. As shown in FIG. 2C, at the sample time T3, the signal intensity is maximum at the azimuth θe and the azimuth θ1, and the signal intensity at the azimuth θe is larger than the signal intensity at the azimuth θ1. Bearing θ
e is a peak based on noise, and the signal direction θ3 is a true peak based on a narrowband signal. Therefore, the signal direction θ
e is incorrectly estimated to be the azimuth of the narrowband signal. FIG. 2 (d)
As shown in (4), at the sample time T4, the signal intensity peaks at the azimuth θ4, and a narrow-band signal arrives from the azimuth θ4.

As shown in FIG. 3, sampling times T1 to T4
, The signal directions θ1 to θ4 are the correct directions of the narrowband signal, but as shown in FIG. 4, at the sample time T3, the direction θe of the signal based on the noise is changed to the sample times T1 and T3.
2 is integrated as the azimuth of the narrowband signal detected. Although the azimuth θ4 of the narrowband signal is correctly detected at the sample time T4, the azimuth θe and the azimuth θ4 at the sample time T3 are too far apart, so that the narrowband signal at the sample time T4 becomes the narrowband signal at the sample times T1 to T3. And the tracking of the narrow band signal is interrupted. The present invention has a problem that the prior art has, as a problem, a peak based on noise is integrated into a narrowband signal being tracked, an erroneous frequency or arrival direction is indicated or tracking is interrupted,
An object of the present invention is to provide a signal tracking method which solves a problem that the signal intensity cannot be used unless the signal intensity is sufficiently large compared to the noise intensity.

[0005]

According to the present invention, in order to solve the above-mentioned problems, a plurality of receiving beams having directivities in different directions are used to monitor a direction in a predetermined continuous range, and each of the receiving directions is monitored. For the received signals respectively received by the wave beams, frequency analysis is performed at each sample time to obtain a signal intensity distribution for the frequency of each received beam,
The following processing is performed in a signal tracking method for continuously measuring the frequency and direction of arrival of a narrowband signal arriving with directionality based on each signal strength distribution. That is,
For each sample frequency, input the vicinity of the sample frequency of each signal intensity distribution to a neural network, the neural network determines whether there is a straight line or curve pattern in the vicinity of the sample frequency, When the neural network determines that the pattern exists, the signal intensity distribution detects that a narrow band signal having a frequency close to the sample frequency has arrived from the azimuth in which the received beam is directed. The frequency having the highest signal strength near the sample frequency of the signal strength distribution is defined as the frequency of arrival of the narrowband signal that has arrived, and the frequency of arrival in the signal strength distribution obtained from a plurality of receiving beams directed near the azimuth. Based on the signal intensity of the narrowband signal, the precise azimuth of the narrowband signal is calculated based on the And the purified comparing measured azimuth and incoming frequency and Seihaka orientation of narrowband signals detected in the past, performs tracking of narrowband signals by integrating the old and new narrowband signal.

[0006]

According to the present invention, since the signal tracking method is configured as described above, the neural network generates a linear or curved pattern in a narrow portion in the vicinity of the sample frequency of each signal intensity distribution for each sample frequency. It is determined whether or not there is, and when the neural network determines that the pattern exists, a narrow band signal having a frequency close to the sample frequency is obtained from the azimuth of the received beam from which the signal intensity distribution is obtained. Detect that it has arrived. Therefore, the above problem can be solved.

[0007]

FIG. 1 is a functional block diagram of a signal tracking device for implementing a signal tracking method according to the present invention. This signal tracking device is an acoustic array sensor 2 that receives a narrow band signal from an unknown sound source 1 having an azimuth θ1 and a frequency f1 by a plurality of receiving beams B1, B2,. Having. On the output side of the acoustic array sensor 2, a frequency for performing a Fourier transform of a received signal of each received beam Bi (i = 1, 2,..., N) to obtain a signal intensity distribution in a frequency space at each sample time The analyzer 3-i is connected. On the output side of each frequency analyzer 3-i, a storage device 4-i for storing a signal intensity distribution in a frequency space for each sample time and an unknown sound source 1 by detecting a signal peak in the signal intensity distribution. The peak detector 6 for detecting the azimuth θ1 and the frequency f1 of the narrowband signal is connected. In the storage device 4-i, a portion near the sample frequency of the signal intensity distribution in the frequency space is extracted, and
A neural network 5-i for determining whether or not the extracted portion has a predetermined pattern based on a narrowband signal from the unknown sound source 1, that is, a linear or curved pattern, is connected thereto. The peak detector 6 is connected to the output side. The output side of the peak detector 6 is connected to a target management device 7 that integrates signals and tracks a narrow band signal.

Next, the signal tracking method of the present invention will be described with reference to FIG. In the acoustic array sensor 2, a narrow band signal from the unknown sound source 1 is received by a plurality of receiving beams B1, B2,.
The received signal of each received beam Bi (i = 1, 2,..., N) is output to the frequency analyzer 3-i. Each frequency analyzer 3-i
Then, a Fourier transform of the received signal of the received beam Bi is performed at each sample time to obtain a signal intensity distribution in a frequency space, and the signal intensity distribution is output to the storage device 4-i and the peak detector 6. The storage device 4-i stores the signal intensity distribution of the received beam Bi for a fixed time in the past. As a result, a signal intensity distribution in the frequency space at each sample time is obtained for each received beam Bi. In each neural network 5-i, for each sample frequency of a plurality of sample frequencies slightly different in frequency, a portion of the signal intensity distribution on the frequency space stored in the storage device 4-i limited to the vicinity of the sample frequency is defined. It is extracted and it is determined whether or not a straight or curved pattern exists in the extracted portion. This is, for example, a pattern of a straight line or a curve is learned in advance, and 1 is set if the learned pattern is present in the extracted input data, and 0 if not.
Can be realized by using a back-propagation type neural network that has been trained to output.
Then, when a portion near the sample frequency f in the signal intensity distribution in the frequency space of the received beam Bi directed to the azimuth θ is input from the storage device 4-i, the neural network 5-i determines that the pattern exists. When it is determined, it is determined that a narrowband signal having a frequency close to the sample frequency f from around the azimuth θ has arrived.

In the case of a narrow band signal, in a signal intensity distribution of a received beam directed in the azimuth of the narrow band signal, a pattern in which points having a large signal intensity are arranged linearly near the frequency of the narrow band signal is formed. Has nature. On the other hand, in the case of a noise signal, a random signal intensity distribution is obtained, and the pattern is not formed. Therefore, the arrival of the narrow-band signal is detected based on whether or not a linear or curved pattern exists in the vicinity of the sample frequency in the signal intensity distribution in the frequency space of the received beam Bi, and is not affected by noise. And a narrow band signal can be detected. If the pattern exists, the neural network 5-i outputs the sample frequency f and the azimuth θ pointing the received beam Bi to the peak detector 6. When it is determined that a narrowband signal having a frequency close to the sample frequency f has arrived from the vicinity of the azimuth θ, the peak detector 6 determines the signal strength in the frequency space obtained from the reception beam Bi directed to the azimuth θ. In the distribution, a frequency f1 having the highest signal strength near the sample frequency f is obtained, and this frequency f1 is set as the arrival frequency of the narrowband signal. Next, interpolation processing is performed on the signal intensity at the frequency f1 of each received beam Bi to obtain the precise measurement azimuth θ1, and the precise measurement azimuth θ1 is determined together with the arrival frequency f1 by the target management device 7.
Output to The precise measurement azimuth θ1 is, for example, a reception beam Bi directed to the azimuth θ and reception beams Bi-1 and Bi on both sides thereof.
It can be obtained by performing a parabolic interpolation using three points on the signal strength of the frequency f1 obtained from +1 and using the method of setting the azimuth at the top of the parabola to the precisely measured azimuth θ1. Up to this point, information that the narrow-band signal of the frequency f1 has arrived from the azimuth θ1 can be obtained. In the target management device 7, the frequency f1 of the narrow band signal detected this time is
And the azimuth θ1 are compared with the frequency and azimuth of the narrowband signal detected in the past, and if the frequency and azimuth are within a predetermined difference, the two signals are determined to be based on the same sound source and integrated.

Thereafter, by repeating the above processing,
It is possible to track a narrow band signal whose frequency and arrival direction change every moment. As described above, this embodiment has the following advantages. (I) In the case of a narrow-band signal, it has the property of forming a pattern in which points having a large signal strength are arranged in a line near the frequency of the narrow-band signal. The signal is detected by detecting this pattern, and the peak of the signal intensity is detected in the vicinity thereof. Therefore, the signal can be tracked without being affected by noise, and the accuracy of the signal tracking can be improved. (Ii) It is possible to track a narrow band signal under a situation where the S / N ratio is lower than that of the conventional method. The present invention is not limited to the above-described embodiment, and various modifications are possible. For example, there are the following modifications. (1) Although the input signal is an audio signal in the above embodiment,
For example, any input signal may be used as long as it has directivity such as electromagnetic waves and can obtain a frequency and a signal strength. (2) The signal tracking device of FIG. 1 may be realized by an individual circuit or a program built in a microcomputer. (3) In the above embodiment, the back propagation type is used as an example of the neural network.
Another neural network such as a Boltzmann machine may be used as long as it can determine the presence of a pattern.

[0011]

As described above in detail, according to the present invention, a portion near the sample frequency of each signal intensity distribution is input to the neural network for each sample frequency, and the neural network generates a signal near the sample frequency. It is determined whether or not a linear or curved pattern exists in the portion, and when the neural network determines that the pattern exists, the signal intensity distribution is calculated from the azimuth in which the received beam is determined to the sample frequency. It detects that a narrow band signal having a close frequency has arrived. Therefore, tracking of a narrow band signal can be performed without being affected by noise, and tracking of a narrow band signal can be performed in a situation where the S / N ratio is low.

[Brief description of the drawings]

FIG. 1 is a functional block diagram of a signal tracking device for implementing a signal tracking method of the present invention.

FIG. 2 is a diagram illustrating a signal intensity distribution in an azimuth space at sampling times T1 to T4 with respect to a frequency f.

FIG. 3 is a diagram illustrating a desirable tracking result of the narrowband signal of FIG. 2;

FIG. 4 is a diagram showing a conventional tracking result of the narrowband signal of FIG. 2;

[Explanation of symbols]

1 Unknown sound source 2 Acoustic array sensor 3-i (i = 1, 2,..., N) Frequency analyzer 4-i (i = 1, 2,..., N) Storage device 5-i (i = 1, 2) , ..., n) Neural network 6 Peak detector 7 Target management device

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Masato Yamashita 1-7-12 Toranomon, Minato-ku, Tokyo Oki Electric Industry Co., Ltd. (72) Inventor Takashi Mizota 1-7-112 Toranomon, Minato-ku, Tokyo Oki Electric Industry Co., Ltd. (72) Inventor Shunji Ozaki 1-7-12 Toranomon, Minato-ku, Tokyo Oki Electric Industry Co., Ltd. (56) References JP 4-9782 (JP, A) JP JP-A-4-136779 (JP, A) JP-A-5-60851 (JP, A) JP-A-3-179285 (JP, A) JP-B-2-26749 (JP, B2)

Claims (1)

(57) [Claims] An azimuth in a predetermined continuous range is monitored using a plurality of receiving beams having directivities in different azimuths, and a sampling time is set for a received signal received by each of the receiving beams. A frequency analysis is performed for each reception beam to determine a signal intensity distribution with respect to the frequency of each received beam.Based on each of the signal intensity distributions, the frequency and arrival direction of the narrowband signal arriving with directionality are temporally continued. In the signal tracking method to be measured, a portion near the sample frequency of each of the signal intensity distributions is input to a neural network for each sample frequency, and the neural network determines whether a linear or curved pattern exists near the sample frequency. The neural network determines that the pattern exists, and determines whether the signal strength It is detected that a narrow band signal having a frequency close to the sample frequency has arrived from the direction in which the received beam is directed, and the frequency having the highest signal intensity in the vicinity of the sample frequency in the signal intensity distribution arrives. Based on the signal strength of the arrival frequency in the signal strength distribution obtained from the plurality of receiving beams directed to the vicinity of the azimuth, the interpolation direction is used to determine the precise azimuth of the narrowband signal. Calculating, comparing the arrival frequency and the precise azimuth with the arrival frequency and the precise azimuth of the narrowband signal detected in the past, and integrating the new and old narrowband signals to track the narrowband signal. Characteristic signal tracking method.