A detection method for effectively reducing the false alarm probability of the guiding signal in underwater acoustic voice communication

Abstract

The present invention discloses a detection method for effectively reducing the false alarm probability of the guiding signal in underwater acoustic voice communication. This detection method is implemented in the following way: an HFM signal with a corresponding frequency is selected as the guiding signal for underwater acoustic voice communication. The receiving end first detects the guiding signal, obtains the corresponding starting edge of the guiding signal through sliding correlation processing, intercepts the guiding signal segment using the starting edge, and performs band-pass downsampling or band-pass under-sampling processing. Then, the guiding signal saved locally is subjected to segmented inner product processing with the obtained signal segment, and the voice guiding signal detection is completed by comparing the segmented inner product value with a preset threshold. The present invention can effectively reduce the false alarm probability of the detected guiding signal, has good detection performance, is applicable to underwater acoustic voice communication, and has very good practicability.

Description

Technical Field:

[0001] The present invention relates to the field of underwater acoustic voice communication, and particularly to a detection method for effectively reducing the false alarm probability of a guiding signal in underwater acoustic voice communication. Background Art:

[0002] With the continuous exploration of the underwater world by humans, the demand for real-time underwater voice communication has become increasingly urgent. Underwater acoustic communication is the main communication method capable of long-distance real-time underwater information interaction. The traditional underwater acoustic voice communication method is to receive and demodulate analog voice communication signals in real time. When no effective voice communication signal is received, noise is continuously demodulated and output, affecting the listening effect and resulting in low work efficiency. To address this problem, currently, the voice activity detection (VAD) method is generally used to distinguish between noise and voice signals. However, voice signals are complex time-varying signals, and the signal feature extraction by this method is complex. The time-domain feature parameters lack good stability and the discrimination ability is poor, with a high false alarm rate. Further, the hyperbolic frequency modulation (HFM) signal is also used in underwater acoustic voice communication as a guiding signal to distinguish actual voice communication signals from noise. The HFM signal is commonly used in underwater acoustic digital communication, has a large time-bandwidth product, and obvious signal features. At the receiving end, the local copy signal can be correlated with the received signal by sliding correlation processing to obtain the correlation value of the guiding signal, and compared with a threshold to determine whether it is a voice communication signal. However, due to the high similarity between the characteristics of human analog voice signals and the frequency spectrum characteristics of frequency modulation signals, false alarm phenomena also occur, affecting the effect of underwater acoustic voice communication. Summary of the Invention:

[0003] The technical problem to be solved by the present invention is to provide a detection method for effectively reducing the false alarm probability of a guiding signal in underwater acoustic voice communication, aiming at the disadvantages of the complex and unstable performance of the voice activity detection method adopted by underwater acoustic voice communication nodes and the easy occurrence of false alarms. By using this method, the false alarm probability of the detected guiding signal can be effectively reduced, the detection performance is good, it is applicable to underwater acoustic voice communication, and has good practicability.

[0004] The technical solution of the present invention is to provide a detection method for effectively reducing the false alarm probability of a guiding signal in underwater acoustic voice communication. This detection method is implemented as follows:

[0005] Select an HFM signal with a corresponding frequency as the guiding signal for underwater acoustic voice communication. The receiving end first detects the guiding signal, obtains the starting edge of the corresponding guiding signal through sliding correlation processing, intercepts the guiding signal segment using the starting edge, and performs band-pass downsampling or band-pass undersampling processing. Then, the locally stored guiding signal is subjected to segmented inner product processing with the obtained signal segment, and the voice guiding signal is detected by comparing the segmented inner product value with a preset threshold.

[0006] Among them, the HFM signal is used as the guiding signal. The common detection method is to perform sliding correlation processing on the received signal directly with the HFM signal saved locally. However, it is sensitive to some voice signals and is prone to false alarms. When intercepting the guiding signal segment, the present invention obtains the corresponding correlation value and the starting edge of the guiding signal through copying the guiding signal for sliding correlation processing. The position of the guiding signal is the starting edge plus the length of the guiding signal, and the guiding signal with the corresponding length is intercepted.

[0007] Preferably, after intercepting the guiding signal with the corresponding length, perform equidistant decimation bandpass downsampling processing on the intercepted guiding signal. If the ratio of the center frequency to the bandwidth of the guiding signal is large, use bandpass undersampling processing to reduce the number of samples and reduce the amount of segment inner product operations in the post-processing.

[0008] Preferably, perform segment inner product processing on the received signal after equidistant decimation bandpass downsampling processing of the intercepted guiding signal with the copied guiding signal saved locally. After the inner product of each segment of the signal, divide by the maximum value respectively for normalization processing to obtain multiple inner product value coefficients, and finally average multiple inner product values to obtain a mean value. The detection of the voice guiding signal is completed by comparing this mean value with a preset threshold.

[0009] Preferably, when detecting the signal, receive the signal in real time. The voice signal containing HFM is the signal to be detected. Use the binary hypothesis detection model to describe the signal perception problem of underwater acoustic voice communication:

[0010]

[0011] Among them, y(n) is the signal to be detected received at the receiving end, s(n) is the voice signal containing the HFM guiding sequence; w(n) is the additive white Gaussian noise in the measured signal and conforms to h(n) and M respectively represent the channel and the number of sampling points; H0 represents that there is no signal to be detected in the current received signal; H1 represents that the signal to be detected is detected.

[0012] Further, when detecting the signal, use f * (n) and g * (n) to represent the energy normalization of the transmitted signal f(n) and the received signal g(n). Use R1(t) to represent the calculation of the correlation degree between the locally copied guiding signal and the signal to be detected, and ρ1 represents the normalized correlation value:

[0013]

[0014]

[0015] ρ1 = max(R1(n))

[0016] And perform sliding correlation processing on the local copy of the pilot signal and the signal to be detected, and compare it with the selected threshold γ1. When ρ1≥γ1, intercept the HFM pilot sequence to be detected using the starting edge obtained by the correlation processing, and enter the segmented inner product detection stage.

[0017] After adopting the above scheme, compared with the prior art, the present invention has the following advantages:

[0018] Based on the sliding correlation detection, the present invention adds a segmented inner product detection method for two-level threshold judgment, further reducing the false alarm probability and improving the underwater acoustic voice communication effect; using this method can effectively reduce the false alarm probability of detecting the pilot signal, with good detection performance, applicable to underwater acoustic voice communication, and having good practicability. Brief Description of the Drawings:

[0019] Figure 1 It is a schematic diagram of the working process of the HFM pilot signal detection method.

[0020] Figure 2 It is the time-frequency diagram of the up-down frequency modulation HFM signal in the measured channel.

[0021] Figure 3 It is the time-frequency diagram of the voice signal in the measured channel. Specific Embodiments:

[0022] The following further describes the present invention in conjunction with the drawings in terms of specific embodiments:

[0023] Figure 1 It is a schematic diagram of the working process of a pilot signal detection method for effectively reducing the false alarm probability applicable to underwater acoustic voice communication nodes proposed by the present invention. The HFM signal has a large time-bandwidth product and is commonly used as a synchronization sequence in digital communication. Figure 2 It is the time-frequency diagram of the up-down frequency modulation HFM signal in the actual test channel. When detecting the signal, the signal is received in real time, and the voice signal containing HFM is the signal to be detected. The binary hypothesis detection model is used to describe the signal perception problem of underwater acoustic voice communication:

[0024]

[0025] Among them, y(n) is the signal to be detected received at the receiving end, s(n) is the voice signal containing the HFM pilot sequence; w(n) is the additive Gaussian white noise in the measured signal and conforms to h(n) and M respectively represent the channel and the number of sampling points; H0 represents that there is no signal to be detected in the current received signal; H1 represents that the signal to be detected is detected and the next signal analysis and processing can be carried out.

[0026] When detecting the signal, use f * (n) and g *(n) represents the energy normalization of the transmitted signal f(n) and the received signal g(n). Let R1(t) denote the calculation of the correlation between the local copy signal and the signal to be detected, and ρ1 denote the normalized correlation value:

[0027]

[0028]

[0029] ρ1 = max(R1(n)) (4)

[0030] The local copy signal and the signal to be detected are subjected to sliding correlation processing and compared with the selected threshold γ1. When ρ1 ≥ γ1, it indicates that the HFM guidance signal to be detected is detected. However, considering that the spectral characteristics of the voice signal are similar to those of the HFM, when the HFM signal to be detected is a partial voice signal similar to the real HFM, the calculated correlation value will also be greater than γ1, and the system is prone to misjudgment and false alarm. At this time, the starting edge obtained by the correlation processing is used to intercept the HFM guidance sequence to be detected, and the segmented inner product detection stage is entered:

[0031] For ease of understanding, we define the transmitted signal vector The intercepted HFM guidance sequence vector to be detected g(n0) represents the signal at the starting edge of the HFM signal to be detected. The locally saved HFM guidance signal segment f 1 and the intercepted HFM guidance sequence g to be detected 1 are both divided into L subsequences with a length of f 1 l 、g 1 l where The corresponding signal segments are respectively subjected to inner product calculation to obtain the inner product value ρ of the subsequence 2l , and finally the normalized segmented inner product value is ρ2.

[0032]

[0033]

[0034] Figure 3 is the time-frequency diagram of the voice signal of the actual measured channel in Fuxian Lake. Since only some frequency components of the voice signal that are similar to the HFM signal characteristics cause false alarms, after the copy correlation processing, the inner product values of the frequency components that are prone to false alarms in the segmented inner product are all attributed to the remaining segmented signal values. If it is not a legal HFM signal, the segmented inner product value ρ2 will not exceed the selected threshold γ2.

[0035] The total false alarm probability can be expressed as

[0036]

[0037] Wherein: and respectively represent the false alarm probabilities of copy-related detection and in-segment inner product detection, and the false alarm probability is reduced by two-level detection thresholds.

[0038] Generally, when the signal bandwidth is fixed, as the sampling rate increases, the signal bandwidth remains unchanged, the energy remains unchanged, the noise introduced by the signal becomes wider, which will cause the in-segment inner product value to fluctuate. Moreover, a large sampling rate will also lead to a large amount of computation. To keep the overall performance of the system consistent and reduce the amount of computation, the received signal can be decimated by equal interval sampling. If the ratio of the signal center frequency to the bandwidth is large, bandpass undersampling technology can also be used to reduce the number of sampling points and the related amount of computation.

[0039] The present invention determines the authenticity of the guiding signal through the sliding correlation processing of the copy signal and the multi-level detection of the in-segment inner product. After the processing results of the sliding correlation and the in-segment inner product both exceed the corresponding threshold values, the actual voice signal will continue to be demodulated, effectively reducing the false alarm probability of the underwater acoustic voice guiding signal and can be preferably used in underwater voice communication nodes.

[0040] The above is only an illustration of the preferred embodiments of the present invention, and it should not be construed as a limitation to the claims. Any equivalent structure or equivalent process transformation made by using the specification of the present invention is included in the patent protection scope of the present invention.

Claims

1. A detection method for effectively reducing the false alarm probability of the guiding signal in underwater acoustic voice communication, characterized in that: This detection method is implemented as follows: Select the HFM signal with the corresponding frequency as the guiding signal for underwater acoustic voice communication. The receiving end first detects the guiding signal, obtains the starting edge of the corresponding guiding signal through sliding correlation processing, intercepts the guiding signal segment using the starting edge, and performs band-pass downsampling or band-pass undersampling processing. Then, perform segmented inner product processing on the locally saved guiding signal and the obtained signal segment, and complete the detection of the voice guiding signal by comparing the segmented inner product value with the preset threshold; Among them, the segmented inner product processing operation is as follows: after the inner product of each segment of the signal, divide by the maximum value respectively for normalization processing to obtain multiple inner product value coefficients, and finally average multiple inner product values to obtain a mean value. Complete the detection of the voice guiding signal by comparing the size of this mean value with the preset threshold.

2. The detection method for effectively reducing the false alarm probability of the underwater acoustic voice communication guidance signal according to claim 1, characterized in that: When intercepting the guiding signal segment, obtain the corresponding correlation value and the starting edge of the guiding signal through copying the guiding signal sliding correlation processing, and intercept the guiding signal of the corresponding length.

3. The detection method for effectively reducing the false alarm probability of the underwater acoustic voice communication guidance signal according to claim 2, wherein: After intercepting the guiding signal of the corresponding length, perform equally-spaced decimation band-pass downsampling processing on the intercepted guiding signal. If the ratio of the center frequency to the bandwidth of the guiding signal is large, use band-pass undersampling processing to reduce the number of sample points.

4. The detection method for effectively reducing the false alarm probability of the underwater acoustic voice communication guiding signal according to claim 3, characterized in that: Perform segmented inner product processing on the received signal after equally-spaced decimation band-pass downsampling processing of the intercepted guiding signal using the locally saved copied guiding signal.

5. The detection method for effectively reducing the false alarm probability of the underwater acoustic voice communication guiding signal according to claim 1, characterized in that: When detecting the signal, receive the signal in real time. The voice signal containing HFM is the signal to be detected. Use the binary hypothesis detection model to describe the signal perception problem of underwater acoustic voice communication: where y(n) is the signal to be detected received at the receiving end, s(n) is the speech signal containing the HFM pilot sequence; w(n) is the additive white Gaussian noise in the measured signal and conforms to h(n) and M represent the channel and the number of sampling points respectively; H0 represents that there is no signal to be detected in the current received signal; H1 represents that the signal to be detected is detected.

6. The detection method for effectively reducing the false alarm probability of the underwater acoustic voice communication guiding signal according to claim 1, wherein: When detecting the signal, use f * (n) and g * (n) to represent the energy normalization of the transmitted signal f(n) and the received signal g(n). Use R1(t) to represent the calculation of the correlation between the local copy of the guiding signal and the signal to be detected, and ρ1 to represent the normalized correlation value: ρ1 = max(R1(n)) And perform sliding correlation processing on the locally copied guiding signal and the signal to be detected and compare it with the selected threshold γ1. When ρ1 ≥ γ1, intercept the HFM guiding sequence to be detected using the starting edge obtained by the correlation processing, and enter the segmented inner product detection stage.

Images