JP2001349940A - Pulse sound detecting system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make certainly detectable a pulse component. SOLUTION: This pulse sound detecting system is provided with a calculating means of an average level of a plurality of bins at a background noise level divided for every frequency bin, a means for detecting a narrow band signal in the background noise level by comparing the average level of the plurality of bins with the background noise level every frequency bin, a calculating means of an average level of a plurality of samples of time series data used for estimating the background noise level, a means for discriminating possibility of a pulse sound based on variation with time of the average level of the plurality of samples, and a means for replacing the background noise level of the frequency bin of which the narrow band signal is detected by the average value of the background noise levels of front and back frequency bins when the narrow band signal is detected and possibility of the pulse sound is high. The means for detecting the narrow band signal judges possibility that continuous sound in the narrow band and the pulse sound in the narrow band leak into the background noise level and discriminates the pulse sound in the narrow band based on the variation of the average level with time, thereby removing the pulse sound taken into the background noise level.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for receiving a pulse sound emitted from a ship or the like by a sensor, estimating a background noise level from a portion of the signal received by the sensor where no pulse sound exists, and The present invention relates to a pulse sound detection system that determines a portion where a pulse sound exists based on a noise level estimation result.

[0002]

2. Description of the Related Art FIG. 2 is a functional block diagram showing one configuration example of a conventional pulse sound detection system.

This pulse sound detection system comprises an input terminal 1, a frequency division processing unit 2, a narrow band instantaneous power calculation processing unit 3, a pulse component detection processing unit 4, a narrow band background noise level estimation processing unit 5, a threshold calculation processing unit. 6 and an output terminal 7.

FIG. 3 is a diagram showing a waveform example of a pulse sound received by a sensor.

The range of arrows described below the waveforms is as follows:
A range in which FFT is performed in a frequency division processing unit described later is shown, and a number described below a range of an arrow is a cycle number in which the present process is performed.

Next, the operation of the conventional pulse sound detection system will be described with reference to FIG.

A signal obtained by digitizing the pulse sound and the background noise as shown in FIG. 3 is input to the input terminal 1, and the signal is output to the frequency division processing unit 2.

The frequency division processing unit 2 applies the signal input from the input terminal 1 to the signal as shown in FIG.
The signal is cut out while overlapping the signal with a certain time width, and an appropriate window function is applied to the cut out signal, and then the input signal is subjected to discrete Fourier transform such as fast Fourier transform (hereinafter, referred to as FFT). The frequency bin is divided into a plurality of frequency bins, and the division result is output to the narrow-band instantaneous power calculation processing unit 3.

In the narrow-band instantaneous power calculation processing unit 3, the result of the frequency division processing unit 2 is input, and the division result is squared using Equation (1) for each frequency bin, thereby obtaining
The narrow-band instantaneous power, which is short-time integrated power, is calculated, and the instantaneous power is output to the pulse component detection processing unit 4 and the narrow-band background noise level estimation processing unit 5.

[0010]

(Equation 1)

Where X (k; n) is the frequency-divided output of the n-th frequency bin in cycle k. P (k; n) is the instantaneous power of the n-th frequency bin in cycle k.

The pulse component detection processing unit 4 receives the instantaneous power from the narrow band instantaneous power calculation processing unit 3 and
A threshold value for pulse component detection is input from the threshold value calculation unit 6, and the instantaneous power is compared with the threshold value for each frequency bin. For instantaneous power exceeding the threshold value, the pulse component (part of the pulse sound portion) (Hereinafter referred to as a pulse component), and outputs the detected instantaneous power and the frequency information of the frequency bin to the output terminal 7.

For example, the result of calculating the instantaneous power in period 4 shown in FIG. 3 is detected as a pulse component. Further, the pulse component detection processing unit 4 outputs a control signal indicating that the estimated value of the background noise level is updated when the pulse component is not detected, and updates the estimated value of the background noise level when the pulse component is detected. A control signal indicating stop is output to the narrow-band background noise level estimation processing unit 5.

In the narrow-band background noise level estimation processing section 5,
The instantaneous power is input from the narrow-band instantaneous power calculation processing unit 3 and the control signal is input from the pulse component detection processing unit 4. If this control signal is a signal indicating an update of the estimated value of the background noise level of the nth frequency bin, an integration process is performed for each frequency bin using the instantaneous power according to Expression (2). Thus, the estimated value of the background noise level is updated, and the result is output to the threshold value calculation processing unit 6. Further, when the control signal is a signal indicating that the update of the estimated value of the background noise level of the nth frequency bin is stopped,
As shown in Expression (3), the estimated value of the background noise level up to the previous time is output to the threshold value calculation processing unit 6.

[0015]

(Equation 2)

Here, α, β: integration constant H (k; n): an estimated value of the background noise level of the nth frequency bin in the period k.

[0017]

(Equation 3)

The threshold value calculation processing unit 6 calculates a threshold value for each frequency bin from the estimated value of the background noise level input from the narrow band background noise level estimation processing unit 5 in consideration of a false alarm probability and the like. Is output to the pulse component detection processing unit 4 as a threshold value used in the pulse component detection processing unit 4.

[0019]

However, the above-mentioned conventional pulse sound detection system has the following problems.

In the conventional pulse sound detection system, for example, when a ship or the like gradually comes close from a distance while continuously emitting the pulse sound, the pulse sound is calculated as the background noise level together with the background noise, and the background noise level is gradually increased. As a result, there is a problem that a pulse sound cannot be detected even when approaching.

The problem of the conventional pulse sound detection system when the pulse sound level gradually rises, which is the above problem, will be specifically described. However, the description of the input terminal 1, the frequency division processing unit 2, and the narrow-band instantaneous power calculation processing unit 3 is omitted because the processing contents do not change.

The pulse component detection processing unit 4 compares the instantaneous power and the threshold value for each frequency bin. If the sound source is sufficiently distant, the level is small even if the instantaneous power includes a pulse sound. And cannot be detected as a pulse component. Therefore, the narrow-band background noise level estimation processing section 5 performs integration processing with the instantaneous power including the pulse sound, and outputs the result to the threshold calculation processing section 6.

Therefore, the threshold value is calculated from the level including the pulse sound.

If these processes are continuously continued as they are, as the pulse sound approaches and the level of the narrow-band instantaneous power increases, the background noise level and the threshold value calculation processing unit in the narrow-band background noise level estimation processing unit 5 will be described. At the same time, the level of the threshold value of 6 also gradually increases. Even if the instantaneous power increases due to the pulse component, the threshold value also increases, and as a result, it cannot be detected as a pulse component.

An object of the present invention is to provide a pulse sound detection system which can eliminate the above-mentioned problems and can reliably detect a pulse component.

[0026]

In order to achieve the above object, the present invention provides: [1] In a pulse sound detection system, a pulse sound emitted from a ship or the like is received by a sensor, and a signal received by the sensor is received. Is divided into frequency bins by a method such as Fourier transform,
The instantaneous power of each frequency bin is calculated from the result of the division into the frequency bins, and the level of the background noise is estimated for each frequency bin from the portion where the pulse sound does not exist in the instantaneous power divided into the frequency bins, Based on the background noise level estimation result, in the pulse sound detection method for determining a portion where a pulse sound is present in the instantaneous power divided into the frequency bins, the background noise level divided for each frequency bin is calculated in the frequency direction. Means for calculating the average level of the plurality of bins while overlapping, and comparing the average level of the plurality of bins with the background noise level divided for each of the frequency bins, thereby obtaining a narrowband signal captured in the background noise level. And time-series data used for estimating the background noise level are overlapped in the time direction. Means for calculating the average level of a plurality of samples, means for determining the possibility of a pulse sound based on a temporal change in the average level of the plurality of samples, a narrow band signal is detected, and the narrow band signal is Means for replacing the background noise level of the frequency bin in which the narrowband signal is detected with an average value of the background noise levels of the frequency bins before and after the narrowband signal when it is determined that the possibility is high. The possibility of a narrow-band continuous sound or a narrow-band pulse sound leaking into the background noise level by means of detecting a signal is determined, and a narrow-band pulse sound is determined based on a temporal change in the average level. By doing so, the pulse sound captured in the background noise level is removed.

[2] In the pulse sound detection system according to the above [1], means for calculating an average level of a plurality of samples while overlapping the time series data used for estimating the background noise level in the time direction; Means for determining the possibility of a pulse sound based on a temporal change in an average level of a plurality of samples, instead of the instantaneous power for each frequency bin used for estimating the background noise level,
Means for calculating an average level of a plurality of bins including the frequency bin in which the narrowband signal is detected, and means for determining the possibility of a pulse sound based on a temporal change in the average level of the instantaneous power. Features.

[0028]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a functional block diagram of a pulse sound detection system according to a first embodiment of the present invention. The input terminal 1, frequency division processing unit 2, and narrow-band instantaneous power calculation processing unit are the same as those of the prior art. 3, a pulse component detection processing unit 4, a threshold calculation processing unit 6, and an output terminal 7 are newly provided with a wideband background noise level estimation processing unit 11, a background noise level comparison processing unit 12, an instantaneous power calculation processing unit 13, an instantaneous power addition. A processing unit 14, an addition level comparison processing unit 15, a background noise level replacement instruction processing unit 16, and a narrow band background noise level estimation processing unit 17 are added.

The broadband background noise level estimation processing unit 11
The integration result for each frequency bin calculated by the narrowband background noise level estimation processing unit 17 is input, and an average value for several bins is calculated as shown in Expression (4) (hereinafter, this result is calculated using the narrowband background noise level). (Referred to as a broadband background noise level because the bandwidth is wider than the integration result for each frequency bin), and outputs the result and the integration result for each frequency bin to the background noise level comparison processing unit 12. .

[0031]

(Equation 4)

Where: L: the number of frequency bins of the narrow-band background noise level used for adding one wide-band background noise level, m = 1,..., N / L J (k; m): the m-th wide band in period k The instantaneous power of

The background noise level comparison processing unit 12 sets a threshold value (hereinafter referred to as threshold value A for discrimination from other threshold values) for each broadband background noise level in consideration of a false alarm probability and the like from the estimated value of the broadband background noise level. The calculated narrowband background noise level, which is the result of integration for each frequency bin, is compared with a threshold A calculated from a band that also includes the narrowband bin.
If there is a narrowband background noise level exceeding the control signal, the control signal is output to the instantaneous power calculation processing unit 13 and
The frequency bin number of the narrowband background noise level exceeding the threshold A and the wideband background noise level of the band including the narrowband background noise level are output to the background noise level replacement instruction processing unit 16.

When the control signal arrives, the instantaneous power calculation processing unit 13 squares each digitized sample signal of the corresponding processing cycle inputted from the input terminal 1 using the equation (5), thereby obtaining the instantaneous power. Is calculated and output to the instantaneous power addition processing unit 14.

[0035]

(Equation 5)

Here, Y (k; t) is the t-th sample signal in cycle k. Q (k; t) is the instantaneous power of the t-th sample signal in cycle k.

The instantaneous power addition processing section 14 adds the instantaneous power of each sample input from the instantaneous power calculation processing section 13 for every several samples as shown in Expression (6), and adds the added power to an addition level comparison processing. Output to the unit 15.

[0038]

(Equation 6)

Where V: the number of samples used for one addition result R (k; u): the u-th addition result in period k.

The addition level comparison processing section 15 calculates the addition power as shown in equation (7), and calculates a change in the addition power level. The change in the addition power is determined by the threshold set for discriminating between the continuous sound and the pulse sound. (Below threshold B to distinguish it from other thresholds
), The control signal is output to the background noise level replacement instruction processing unit 16.

C (k; u) = R (k; u) / R (k; u−1) (7) where C (k; u) is the u-th level change in period k.

When a control signal is received, the background noise level replacement instruction processor 16 determines the frequency bin number of the narrow band background noise level exceeding the threshold value A and the wide band background noise level of the band including the narrow band background noise level. Output to the narrowband background noise level estimation processing unit 17.

In the narrow-band background noise level estimation processing section 17, the frequency bin number of the narrow-band background noise level and the wide-band background noise level of the band including the narrow-band background noise level are transmitted from the background noise level replacement instruction processing section 16. In this case, the narrow-band background noise level of the narrow-band frequency bin is replaced with the wide-band background noise level of the band including the narrow-band background noise level, and the replaced narrow-band background noise level is output to the threshold value calculation processing unit 6.

The threshold value calculation processing section 6 calculates a threshold value from the narrow band background noise level replaced by the above conditions input from the narrow band background noise level estimation processing section 17.

Next, a second embodiment of the present invention will be described.

FIG. 4 is a functional block diagram of a pulse sound detecting system according to a second embodiment of the present invention. The difference from the first embodiment is that the data input to the instantaneous power level comparison processing unit 22 is changed to the narrow band instant power for each frequency bin, which is the output of the narrow band instant power calculation processing unit 3.

This is effective when the pulse width and pulse repetition period of the pulse sound to be identified are sufficiently longer than the execution range of one discrete Fourier transform such as FFT, and the background noise level comparison processing unit 21 exceeds the threshold A. When the narrow-band background noise level exists, the control signal and the frequency bin number of the narrow-band background noise level exceeding the threshold value A (hereinafter, referred to as the corresponding frequency bin number) are output to the instantaneous power level comparison processing unit 22. , And outputs the wideband background noise level of the band including the corresponding frequency bin number and the narrowband background noise level to the background noise level replacement instruction processing unit 16.

When the control signal and the corresponding frequency bin number arrive, the instantaneous power level comparison processing unit 22 calculates not the instantaneous power including the entire band but the narrow-band instantaneous power of the corresponding frequency bin as shown in equation (8). A comparison is made with the narrow-band instantaneous power of the corresponding frequency bin in the previous cycle, and when the level change as the comparison result exceeds the threshold B, a control signal is output to the background noise level replacement instruction processing unit 16.

D (k; n) = P (k; n) / P (k−1; n) (8) where D (k; n) is the level change of the n-th frequency bin in period k. is there.

Input terminal 1, frequency division processing unit 2, narrow-band instantaneous power calculation processing unit 3, pulse component detection processing unit 4, threshold calculation processing unit 6, output terminal 7, wide-band background noise level estimation processing unit 11, background noise The level replacement instruction processing unit 16 and the narrow band background noise level estimation processing unit 17 perform the same processing as in the first embodiment.

The present invention has the following utilization modes.

In the first and second embodiments, the wideband background noise level calculated by the wideband background noise level estimation processing section 11 is obtained by dividing the narrowband background noise level for each band to be added, Although an example in which calculation is performed without overlap is shown, a narrowband background noise level to be added may be overlapped to calculate a wideband background noise level.

In the first embodiment, the instantaneous power addition processor 1
4 shows an example in which the added power calculated in step 4 is divided into several samples to be added and calculated without overlapping the instantaneous power of each sample. However, the added power is calculated by overlapping the added instantaneous power. No problem.

In the first and second embodiments, when the background noise level comparison processing section 12 (in the second embodiment, the background noise level comparison processing section 21) outputs a control signal, the instantaneous power calculation processing section 13, Although an example is shown in which the power addition processing unit 14 and the addition level comparison processing unit 15 (in the second embodiment, the instantaneous power level comparison processing unit 22) are performed,
Instantaneous power calculation processing unit 13, instantaneous power addition processing unit 14
The addition level comparison processing unit 15 also performs processing in parallel at the same time, and the control signal is output from both the background noise level comparison processing unit 12 (21) and the addition level comparison processing unit 15 to the background noise level replacement instruction processing unit 16. In some cases, a replacement instruction of the narrowband background noise level and the wideband background noise level may be issued.

In the second embodiment, the frequency bins to be compared by the instantaneous power level comparison processing unit 22 have been described as a single frequency bin. However, the comparison process may be performed using a plurality of frequency bins added in the frequency direction.

In the first embodiment, the addition level comparison processing section 15
, The addition result R (k; u) is replaced by the immediately preceding R (k; u−
1), but R (k; u-2) or R (k; u-
As x), a transitional portion where a pulse appears can be removed to make a comparison.

Similarly, in the second embodiment, the instantaneous power level comparison processing section 22 compares the instantaneous power P (k; n) with P (k-1; n) in the immediately preceding cycle. 2;
n) or P (k−x; n) can be compared to remove the transient part where the pulse appears.

The device of the above embodiment may be constituted by an individual circuit using an integrated circuit, or may be constituted by software using a digital signal processor (DSP) or the like.

The present invention is not limited to the above-described embodiment, and various modifications can be made based on the gist of the present invention, and these are not excluded from the scope of the present invention.

[0060]

As described above, according to the present invention, the following effects can be obtained.

(1) By comparing the narrow-band background noise level and the wide-band background noise level, if the level of a specific narrow-band bin becomes higher than the level of surrounding bins, there is a possibility that a continuous sound or a pulse sound exists. , The level change is examined by comparing the addition levels, and if there is a sudden level change, it is assumed that a pulse sound may exist, and the narrow-band background noise level is changed to the wide-band background noise level. , The threshold value can be lowered, and even if the narrow-band background noise level gradually rises, it can be detected as a pulse sound if the level of the pulse sound exceeds a certain level.

(2) By using the narrow-band instantaneous power in units of frequency bins for the addition level comparison processing for distinguishing between continuous sounds and pulse sounds, extra band components are cut off and the S / N at the time of level comparison is improved. This makes it easier to detect the pulse sound.

[Brief description of the drawings]

FIG. 1 is a functional block diagram of a pulse sound detection system according to a first embodiment of the present invention.

FIG. 2 is a functional block diagram showing a configuration example of a conventional pulse sound detection system.

FIG. 3 is a diagram illustrating a waveform example of a pulse sound received by a sensor.

FIG. 4 is a functional block diagram of a pulse sound detection system according to a second embodiment of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Input terminal 2 Frequency division processing part 3 Narrow band instantaneous power calculation processing part 4 Pulse component detection processing part 6 Threshold calculation processing part 7 Output terminal 11 Broadband background noise level estimation processing part 12, 21 Background noise level comparison processing part 13 Instantaneous power Calculation processing unit 14 Instantaneous power addition processing unit 15 Addition level comparison processing unit 16 Background noise level replacement instruction processing unit 17 Narrow band background noise level estimation processing unit 22 Instantaneous power level comparison processing unit

Claims (2)

[Claims] 1. A pulse sound emitted from a ship or the like is received by a sensor, a signal received by the sensor is divided into frequency bins by a method such as Fourier transform, and the frequency bin is divided into frequency bins. The instantaneous power of the frequency bins is calculated, the level of the background noise is estimated for each frequency bin from the portion where the pulse sound does not exist in the instantaneous power divided into the frequency bins, and based on the estimation result of the background noise level, In a pulse sound detection system that determines a portion where a pulse sound exists in instantaneous power divided into bins, an average level of a plurality of bins is calculated while overlapping background noise levels divided for each frequency bin in a frequency direction. Means for comparing the average level of the plurality of bins with the background noise level divided for each of the frequency bins. Means for detecting a narrowband signal captured in a level, means for calculating an average level of a plurality of samples while overlapping time-series data used for estimating the background noise level in a time direction, and Means for determining the possibility of a pulse sound based on a temporal change in an average level; and a narrow band signal is detected. When the narrow band signal is determined to have a high possibility of a pulse sound, the narrow band signal is Means for replacing the background noise level of the detected frequency bin with the average value of the background noise levels of the frequency bins before and after the detected frequency bin, wherein the means for detecting the narrowband signal has a narrow-band continuous sound within the background noise level. Or the possibility that the narrow-band pulse sound has leaked is determined, and the temporal change in the average level is determined to be the narrow-band pulse sound. Pulse sound detection system and removing the captured pulse sound noise levels. 2. The pulse sound detection system according to claim 1, wherein said means for calculating an average level of a plurality of samples while overlapping time-series data used for estimating said background noise level in a time direction, and said plurality of samples. Instead of the means for determining the possibility of a pulse sound based on a temporal change in the average level, the instantaneous power of each frequency bin used for estimating the background noise level includes a frequency bin in which the narrowband signal is detected. A pulse sound detection system comprising: means for calculating an average level of a plurality of bins; and means for determining the possibility of a pulse sound based on a temporal change in the average level of the instantaneous power.