Statistical characteristic analysis method based on acoustic signal characteristics of fish schools cultured in large cages

Abstract

The present invention discloses a statistical characteristic analysis method based on the acoustic signal characteristics of fish schools cultured in large cages, characterized in that: the statistical characteristic analysis method is as follows: Step 1, data acquisition: based on a large semi-submersible culture platform, hydrophones are deployed; Step 2, test layout: in order to more clearly analyze the impact of external excitation on fish school noise during the test process, track and record events that may affect the noise signal during the data acquisition process; Step 3, analyze the data; Step 4, results and analysis; The present invention uses signal processing technology to determine the frequency spectrum characteristics of fish. The short-term characteristics of the vocalization frequency bands of different fish schools are different, which can be used to distinguish the vocalization characteristics of different fish in cages; The present invention is extended to the application of passive acoustic technology to analyze the correlation between marine fish behavior and acoustic ecological environment.

Description

Technical Field

[0001] The present invention relates to the technical field of characteristic research of acoustic signals of large-scale cage-cultured fish schools, and in particular to a statistical characteristic analysis method based on acoustic signal characteristics of large-scale cage-cultured fish schools. Background Art

[0002] Passive sonar technology has gradually become an important means and tool for marine fishery resource assessment, conservation and marine ecosystem monitoring. Compared with traditional fishery resource survey methods, acoustic monitoring methods combine the assessment and calculation of fishery resources with scientific and technological means, and have the advantages of high accuracy, wide survey range, and all-weather detection. For a long time, people have conducted a lot of investigations and studies on the biological significance of fish vocalization behavior. The structure of fish vocal organs is diverse, and their vocalization mechanisms and vocalization spectrum characteristics are also different. The generally accepted classification methods of fish vocalization are: (1) friction vocalization: when fish eat, their teeth rub against each other to make sounds, and when fish are frightened, their body parts rub and slap to make sounds. The sound frequency generated by the friction mechanism is concentrated between 100 and 8000 Hz, and its main energy distribution is between 1000 and 4000 Hz. Friction sound is a non-harmonic pulse signal with a wide spectrum; (2) vibration vocalization: sound is made by the contraction of vocal muscles to drive the swim bladder to vibrate. The frequency bandwidth of the swim bladder's vibration sound is 50 to 1500 Hz, and there are both fundamental frequency and high-frequency harmonics; (3) Hydrodynamic sound: usually occurs when the fish turns quickly or changes speed quickly. The frequency of these sounds is relatively low, in the infrasound band below 20 Hz. These sound waves are generated when the fish are active and do not contain information about the communication between the population. Existing studies have shown that biological noise with a frequency of 50 Hz to 10 kHz includes swimming noise, biological sound, and feeding noise, which can fully reflect the status of fish schools such as information exchange, defense, and foraging.

[0003] The existing technology is unable to statistically characterize and perform corresponding analysis on the acoustic signal characteristics of large cage cultured fish schools.

[0004] Therefore, the statistical characteristic analysis method based on the acoustic signal characteristics of large-scale cage-cultured fish schools has become an urgent problem to be solved. Summary of the invention

[0005] The technical problem to be solved by the present invention is that the prior art is unable to perform statistical features and corresponding analysis on the acoustic signal features of large cage cultured fish schools.

[0006] In order to solve the above technical problems, the technical solution provided by the present invention is: a statistical characteristic analysis method based on the acoustic signal characteristics of large cage cultured fish schools, and the statistical characteristic analysis method is as follows:

[0007] Step 1: Data collection: Deploy hydrophones based on a large semi-submersible aquaculture platform;

[0008] Step 2, test layout: In order to more clearly analyze the impact of external excitation on the fish school noise during the test process, track and record the events that may affect the noise signal during the data acquisition process;

[0009] Step 3: Analyze the data: Use short-time analysis technology to study the characteristics of noise signals of different fish schools in the cage. Short-time energy describes the energy of each frame of sound signal. Its changing trend is consistent with the waveform of the sound signal, which can well reflect the time domain information of the sound.

[0010] Assume that the time domain signal is x, add the Hanning window to process the frame and get the nth frame signal as x n (m), then x n (m) satisfies the following formula:

[0011] x n (m)=w(m)x(n+m)0≤m≤N-10≤m≤N-1 (1)

[0012] The window function expression of the Hanning window used is:

[0013]

[0014] Where n = 0, 1T, 2T, ..., NT is the frame length, and the sound signal x of the nth frame is n The short-time energy of (m) is expressed as E n The calculation formula is as follows:

[0015]

[0016] The short-time zero-crossing rate refers to the number of times the signal passes through the zero axis in each frame. The zero-crossing rate reflects the frequency information of the signal to a certain extent, and the expression is:

[0017]

[0018] Where sgn[] is the sign function;

[0019] Since the sound is short-term stable in the time period of 10-30ms, the commonly used short-term analysis method takes 10-30ms as the frame length, and the length of each frame signal used for data processing is 25ms, and the frame shift is 10ms;

[0020] Step 4. Results and analysis: Test signal time domain and frequency domain characteristics: In order to obtain the fish school noise characteristics, the data of the entire period is extracted every 5 minutes, the length of each data segment is 6326400, and the duration of each sample segment is about 300s;

[0021] The response characteristics of fish schools to events: During the artificial feeding period, at the feeding point of the grouper cage, the fish schools are more active and the noise intensity is higher than the background noise;

[0022] Automatic feeding machines are used for large-scale feeding. Fish are most active because they are competing for food, and the sound energy they produce is the strongest, which is much higher than the background noise. By playing back the signals of the corresponding period in audio form, the types of sounds heard are mainly flapping, gurgling, gurgling and chewing.

[0023] The platform wave energy continues to generate electricity, and at the same time, small fishing boats pass by. Due to the stimulation of external interference, the noise energy during this period is relatively strong compared to the ocean background noise energy. It can be seen that the activity of the fish school will be increased by the influence of motor power generation and the background noise of small fishing boats;

[0024] At other times, especially at night, the activity of fish schools is low, and there is only swimming noise and surface wave noise generated by sporadic swimming of fish schools. The signals collected by the hydrophone are converted into audio signals for playback. The sounds heard at most times in the middle of the night are mainly natural environmental noises such as flapping and rustling.

[0025] Time-frequency characteristics analysis of artificial feeding: Artificial feeding was carried out in the grouper cages, and the sound signal at that time was selected for time-frequency analysis to obtain the short-time spectrum and power spectrum of that period. In order to obtain the characteristics of the grouper group's feeding and vocalization stage, the noise frequency band of its vocalization was band-pass filtered with a filter bandwidth of 200-1000Hz, and its short-time energy and short-time zero-crossing rate were solved. It can be seen that the short-time energy of the signal during the grouper group's vocalization period is higher than that of other periods, and the peak energy is about 8×10 -4 , the short-term zero-crossing rate is relatively stable compared with other moments;

[0026] Analysis of time-frequency characteristics during the feeding period of the automatic feeding machine: The automatic feeding machine feeds large quantities of food between the two cages, and the sound signal at this moment is selected for time-frequency analysis to obtain the time-frequency diagram of this moment. During this period, there is aliasing of fish school sound signals. Due to the large-scale feeding of the automatic feeding machine, the fish school in the entire cage moves violently, and the fish school is most active. 2000-5000Hz is the frequency band where the fish school hits each other and splashes water, which has higher energy than manual feeding. When the audio signal of this period is played, the hissing and creaking friction sounds can be heard, as well as the sound of the fish swallowing accompanied by the bursting of bubbles on the water surface. Sound; In order to distinguish the two frequency bands of fish school signals that are aliased when the fish school is automatically fed on a large scale, a bandpass filter with a bandwidth of 600-2000Hz is used to filter the time domain signal, extract the time period when the golden pomfret noise appears, and calculate its short-time energy and short-time zero-crossing rate; a bandpass filter with a bandwidth of 200-7000Hz is used to filter the time domain signal, extract the time period when the croaker fish makes sounds, and calculate its short-time energy and short-time zero-crossing rate; the probability statistical distribution of the short-time energy and zero-crossing rate of the golden pomfret and the croaker fish is solved respectively, and the short-time characteristic differences of different fish schools are observed;

[0027] Analysis of time-frequency characteristics during the midnight period: At midnight, the activity of fish schools is low, and the frequency band of the noise they make is also low. There is only radiation noise from sporadic swimming and splashing of water between 2000-4000Hz, and the peak sound pressure level is about 90dB;

[0028] Comparison of vocal characteristics of different fish groups: Based on the above content, the vocalization and short-time characteristics of the three fish groups are compared. It can be seen that the vocalization mechanisms of the three fish groups are different, and their short-time energy and short-time zero-crossing rate characteristics are also different. The probability peak values ​​of short-time energy and short-time zero-crossing rate of the vocal signal of the caller fish are the largest. This is because the caller fish relies on the compression of the vocal muscles to drive the peritoneum to achieve vibration sound, which is mainly in the form of pulse trains and has high-order harmonics. The second is the golden pomfret, which makes sound by rubbing and slapping the tail and fins to produce a relatively wide-band signal form; grouper makes sounds by preying, chewing and swallowing food, with a relatively low frequency, and the probability peak values ​​of short-time energy and short-time zero-crossing rate are the smallest.

[0029] Furthermore, the requirement for deploying the hydrophones described in step 1 is: the hydrophones are placed deep in the cages outside the grouper breeding area, the golden pomfret breeding area and the automatic feed delivery area, and the water depth is 12.5m.

[0030] Furthermore, the hydrophone is a self-contained hydrophone, the model of the hydrophone is TC4037, the sensitivity is -192dB, the amplification factor is 40dB, and during the test, the system sampling rate of the hydrophone is 105440Hz.

[0031] The advantages of the present invention compared with the prior art are: the present invention tests and analyzes the time-frequency domain characteristics of large-scale cage cultured fish within 24 hours, and selects the acoustic signals at three special times: manual feeding, automatic feeding machine feeding, and midnight for time-frequency analysis. First, the original signal is subjected to short-time Fourier transform to obtain the variation law of underwater noise frequency and signal strength. According to the background noise of the ocean and the vocal characteristics of the fish, filters with different bandwidths are used to obtain fish acoustic signals of different frequency bands, and their short-time energy, short-time zero-crossing rate and power spectrum are calculated to compare and distinguish the vocal frequency bands, vocalization methods and short-time characteristics of the three fish. The signal processing technology can be used to determine the spectral characteristics of fish. The vocal frequency bands of different fish schools have different short-time characteristics, which can be used to distinguish the vocal characteristics of different fish in the cage.

[0032] The present invention shows that: (1) feeding activities and external stimuli (such as the passing of fishing boats, power generation, etc.) will cause changes in the behavior and state of fish schools, and their acoustic characteristics can reflect their behavioral information. It can be seen from the power spectrum analysis of underwater acoustic signals that in the range of 100 to 5000 Hz, the energy of the acoustic signals emitted by fish schools is significantly higher than the background noise. The low-frequency background noise of the ocean is within 100 Hz, and the noise intensity gradually decreases and tends to be stable as the frequency increases after 5 kHz. It is mainly composed of the high-frequency background noise of the ocean.

[0033] (2) During the feeding stage of the fish in the cage, the frequency band of the sound emitted by the grouper school is mainly between 200 and 1000 Hz, the frequency band of the sound emitted by the golden pomfret is between 600 and 2000 Hz, and the frequency band of the noise generated by the fish slapping the water due to the competition for food is between 2000 and 5000 Hz. The sound signal of the croaker has a certain periodicity and high-frequency harmonics, with a cut-off frequency of about 7000 Hz. There is a certain frequency overlap in the sound of the golden pomfret and the croaker.

[0034] (3) The peak sound pressure level distribution of different fish schools is different. During the feeding stage, the peak sound pressure level of grouper is about 94dB, the peak sound pressure level of golden pomfret is about 108dB, and the peak sound pressure level of croaker is about 120dB.

[0035] (4) The frequency of the vocal signal of grouper is low, the short-term zero-crossing rate is low and the relative change is relatively stable; the vocal signal of golden pomfret is a non-harmonic pulse signal with a wide spectrum and a high probability distribution dispersion of the short-term zero-crossing rate; the vocal signal of croaker is in the form of a pulse train in the time domain and has the highest short-term energy and short-term zero-crossing rate.

[0036] The present invention is applied to the analysis of the correlation between the behavior of marine fish and the acoustic ecological environment by using passive acoustic technology; the present invention is reasonably designed and is worthy of being vigorously promoted. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1The invention is based on the statistical characteristic analysis method of the acoustic signal characteristics of large-scale cage cultured fish schools to test key time period events.

[0038] Figure 2 The present invention is based on the statistical characteristic analysis method of the acoustic signal characteristics of large-scale cage cultured fish schools, which includes the all-day time domain characteristics (a) and the frequency domain characteristics (b).

[0039] Figure 3 The short-time spectrum (a) and the periodogram power spectrum (b) of the artificial feeding period in the statistical characteristic analysis method of the acoustic signal characteristics of fish schools cultured in large cages according to the present invention.

[0040] Figure 4 The present invention shows the short-time characteristics (a) and the local amplification results (b) of grouper in the statistical characteristic analysis method of the acoustic signal characteristics of fish schools cultured in large cages.

[0041] Figure 5 The short-time energy probability density distribution (a) and the short-time zero-crossing rate probability density distribution (b) of the grouper sound signal in the statistical characteristic analysis method based on the sound signal characteristics of large cage culture fish schools of the present invention are as follows:

[0042] Figure 6 The short-time spectrum (a) and the periodogram power spectrum (b) of the automatic feeding period in the statistical characteristic analysis method based on the acoustic signal characteristics of large-scale cage culture fish schools of the present invention.

[0043] Figure 7 The invention discloses a short-time spectrum (a) and a short-time characteristic (b) of a sound signal of a golden pomfret in a statistical characteristic analysis method of a sound signal characteristic of a school of fish cultured in a large net cage.

[0044] Figure 8 The present invention discloses a short-time spectrum (a) and a short-time characteristic (b) of the vocal signal of a croaker fish in a statistical characteristic analysis method of the acoustic signal characteristics of a school of fish cultured in a large net cage.

[0045] Fig. 9 The invention discloses a short-time energy probability density distribution (a) and a short-time zero-crossing rate probability density distribution (b) of a sound signal of a golden pomfret in a statistical characteristic analysis method based on the characteristics of sound signals of a school of fish cultured in a large net cage.

[0046] Fig.10 The present invention discloses a short-time energy probability density distribution (a) and a short-time zero-crossing rate probability density distribution (b) of the vocal signal of a croaker fish in a statistical characteristic analysis method based on the acoustic signal characteristics of a school of fish cultured in a large net cage.

[0047] Fig.11 The short-time spectrum (a) and the periodogram power spectrum (b) of the midnight period in the statistical characteristic analysis method based on the acoustic signal characteristics of large-scale cage cultured fish schools of the present invention.

[0048] Fig.12 These are three types of fish sound characteristics in the statistical characteristic analysis method of the sound signal characteristics of fish schools cultured in large cages in the present invention. DETAILED DESCRIPTION

[0049] The statistical characteristic analysis method based on the acoustic signal characteristics of large-scale cage cultured fish schools of the present invention is further described in detail below in conjunction with the accompanying drawings.

[0050] Combined with Figure 1-12 , the present invention is introduced in detail.

[0051] A statistical characteristic analysis method based on the acoustic signal characteristics of large cage cultured fish schools is as follows:

[0052] Step 1: Data collection: Deploy hydrophones based on a large semi-submersible aquaculture platform;

[0053] Step 2, test layout: In order to more clearly analyze the impact of external excitation on the fish school noise during the test process, track and record the events that may affect the noise signal during the data acquisition process;

[0054] Step 3: Analyze the data: Use short-time analysis technology to study the characteristics of noise signals of different fish schools in the cage. Short-time energy describes the energy of each frame of sound signal. Its changing trend is consistent with the waveform of the sound signal, which can well reflect the time domain information of the sound.

[0055] Assume that the time domain signal is x, add the Hanning window to process the frame and get the nth frame signal as x n (m), then x n (m) satisfies the following formula:

[0056] x n (m)=w(m)x(n+m)0≤m≤N-10≤m≤N-1 (1)

[0057] The window function expression of the Hanning window used is:

[0058]

[0059] Where n = 0, 1T, 2T, ..., NT is the frame length, and the sound signal x of the nth frame is n The short-time energy of (m) is expressed as E n The calculation formula is as follows:

[0060]

[0061] The short-time zero-crossing rate refers to the number of times the signal passes through the zero axis in each frame. The zero-crossing rate reflects the frequency information of the signal to a certain extent, and the expression is:

[0062]

[0063] Where sgn[] is the sign function;

[0064] Since the sound is short-term stable in the time period of 10-30ms, the commonly used short-term analysis method takes 10-30ms as the frame length, and the length of each frame signal used for data processing is 25ms, and the frame shift is 10ms;

[0065] Step 4. Results and analysis: Test signal time domain and frequency domain characteristics: In order to obtain the fish school noise characteristics, the data of the entire period is extracted every 5 minutes, the length of each data segment is 6326400, and the duration of each sample segment is about 300s;

[0066] The response characteristics of fish schools to events: During the artificial feeding period, at the feeding point of the grouper cage, the fish schools are more active and the noise intensity is higher than the background noise;

[0067] Automatic feeding machines are used for large-scale feeding. Fish are most active because they are competing for food, and the sound energy they produce is the strongest, which is much higher than the background noise. By playing back the signals of the corresponding period in audio form, the types of sounds heard are mainly flapping, gurgling, gurgling and chewing.

[0068] The platform wave energy continues to generate electricity, and at the same time, small fishing boats pass by. Due to the stimulation of external interference, the noise energy during this period is relatively strong compared to the ocean background noise energy. It can be seen that the activity of the fish school will be increased by the influence of motor power generation and the background noise of small fishing boats;

[0069] At other times, especially at night, the activity of fish schools is low, and there is only swimming noise and surface wave noise generated by sporadic swimming of fish schools. The signals collected by the hydrophone are converted into audio signals for playback. The sounds heard at most times in the middle of the night are mainly natural environmental noises such as flapping and rustling.

[0070] Time-frequency characteristics analysis of artificial feeding: Artificial feeding was carried out in the grouper cages, and the sound signal at that time was selected for time-frequency analysis to obtain the short-time spectrum and power spectrum of that period. In order to obtain the characteristics of the grouper group's feeding and vocalization stage, the noise frequency band of its vocalization was band-pass filtered with a filter bandwidth of 200-1000Hz, and its short-time energy and short-time zero-crossing rate were solved. It can be seen that the short-time energy of the signal during the grouper group's vocalization period is higher than that of other periods, and the peak energy is about 8×10 -4 , the short-term zero-crossing rate is relatively stable compared with other moments;

[0071] Analysis of time-frequency characteristics during the feeding period of the automatic feeding machine: The automatic feeding machine feeds large quantities of food between the two cages, and the sound signal at this moment is selected for time-frequency analysis to obtain the time-frequency diagram of this moment. During this period, there is aliasing of fish school sound signals. Due to the large-scale feeding of the automatic feeding machine, the fish school in the entire cage moves violently, and the fish school is most active. 2000-5000Hz is the frequency band where the fish school hits each other and splashes water, which has higher energy than manual feeding. When the audio signal of this period is played, the hissing and creaking friction sounds can be heard, as well as the sound of the fish swallowing accompanied by the bursting of bubbles on the water surface. Sound; In order to distinguish the two frequency bands of fish school signals that are aliased when the fish school is automatically fed on a large scale, a bandpass filter with a bandwidth of 600-2000Hz is used to filter the time domain signal, extract the time period when the golden pomfret noise appears, and calculate its short-time energy and short-time zero-crossing rate; a bandpass filter with a bandwidth of 200-7000Hz is used to filter the time domain signal, extract the time period when the croaker fish makes sounds, and calculate its short-time energy and short-time zero-crossing rate; the probability statistical distribution of the short-time energy and zero-crossing rate of the golden pomfret and the croaker fish is solved respectively, and the short-time characteristic differences of different fish schools are observed;

[0072] Analysis of time-frequency characteristics during the midnight period: At midnight, the activity of fish schools is low, and the frequency band of the noise they make is also low. There is only radiation noise from sporadic swimming and splashing of water between 2000-4000Hz, and the peak sound pressure level is about 90dB;

[0073] Comparison of vocal characteristics of different fish groups: Based on the above content, the vocalization and short-time characteristics of the three fish groups are compared. It can be seen that the vocalization mechanisms of the three fish groups are different, and their short-time energy and short-time zero-crossing rate characteristics are also different. The probability peak values ​​of short-time energy and short-time zero-crossing rate of the vocal signal of the caller fish are the largest. This is because the caller fish relies on the compression of the vocal muscles to drive the peritoneum to achieve vibration sound, which is mainly in the form of pulse trains and has high-order harmonics. The second is the golden pomfret, which makes sound by rubbing and slapping the tail and fins to produce a relatively wide-band signal form; grouper makes sounds by preying, chewing and swallowing food, with a relatively low frequency, and the probability peak values ​​of short-time energy and short-time zero-crossing rate are the smallest.

[0074] Requirements for deploying hydrophones as described in step 1: Place the hydrophones deep inside the cages outside the grouper culture area, golden pomfret culture area and automatic feed delivery area, with a water depth of 12.5m.

[0075] The hydrophone is a self-contained hydrophone, the model of the hydrophone is TC4037, the sensitivity is -192dB, the amplification factor is 40dB, and during the test, the system sampling rate of the hydrophone is 105440Hz.

[0076] The specific implementation process of the statistical characteristic analysis method based on the acoustic signal characteristics of large-scale cage cultured fish schools of the present invention is as follows: Step 1, data collection: Based on a large semi-submersible culture platform, a hydrophone is deployed;

[0077] Step 2, test layout: In order to more clearly analyze the impact of external excitation on the fish school noise during the test process, track and record the events that may affect the noise signal during the data acquisition process;

[0078] Step 3: Analyze the data: Use short-time analysis technology to study the characteristics of noise signals of different fish schools in the cage. Short-time energy describes the energy of each frame of sound signal. Its changing trend is consistent with the waveform of the sound signal, which can well reflect the time domain information of the sound.

[0079] Assume that the time domain signal is x, add the Hanning window to process the frame and get the nth frame signal as x n (m), then x n (m) satisfies the following formula:

[0080] x n (m)=w(m)x(n+m)0≤m≤N-10≤m≤N-1 (1)

[0081] The window function expression of the Hanning window used is:

[0082]

[0083] Where n = 0, 1T, 2T, ..., NT is the frame length, and the sound signal x of the nth frame is n The short-time energy of (m) is expressed as E n The calculation formula is as follows:

[0084]

[0085] The short-time zero-crossing rate refers to the number of times the signal passes through the zero axis in each frame. The zero-crossing rate reflects the frequency information of the signal to a certain extent, and the expression is:

[0086]

[0087] Where sgn[] is the sign function;

[0088] Since the sound is short-term stable in the time period of 10-30ms, the commonly used short-term analysis method takes 10-30ms as the frame length, and the length of each frame signal used for data processing is 25ms, and the frame shift is 10ms;

[0089] Step 4. Results and analysis: Test signal time domain and frequency domain characteristics: In order to obtain the fish school noise characteristics, the data of the entire period is extracted every 5 minutes, the length of each data segment is 6326400, and the duration of each sample segment is about 300s;

[0090] The response characteristics of fish schools to events: During the artificial feeding period, at the feeding point of the grouper cage, the fish schools are more active and the noise intensity is higher than the background noise;

[0091] Automatic feeding machines are used for large-scale feeding. Fish are most active because they are competing for food, and the sound energy they produce is the strongest, which is much higher than the background noise. By playing back the signals of the corresponding period in audio form, the types of sounds heard are mainly flapping, gurgling, gurgling and chewing.

[0092] The platform wave energy continues to generate electricity, and at the same time, small fishing boats pass by. Due to the stimulation of external interference, the noise energy during this period is relatively strong compared to the ocean background noise energy. It can be seen that the activity of the fish school will be increased by the influence of motor power generation and the background noise of small fishing boats;

[0093] At other times, especially at night, the activity of fish schools is low, and there is only swimming noise and surface wave noise generated by sporadic swimming of fish schools. The signals collected by the hydrophone are converted into audio signals for playback. The sounds heard at most times in the middle of the night are mainly natural environmental noises such as flapping and rustling.

[0094] Time-frequency characteristics analysis of artificial feeding: Artificial feeding was carried out in the grouper cages, and the sound signal at that time was selected for time-frequency analysis to obtain the short-time spectrum and power spectrum of that period. In order to obtain the characteristics of the grouper group's feeding and vocalization stage, the noise frequency band of its vocalization was band-pass filtered with a filter bandwidth of 200-1000Hz, and its short-time energy and short-time zero-crossing rate were solved. It can be seen that the short-time energy of the signal during the grouper group's vocalization period is higher than that of other periods, and the peak energy is about 8×10 -4 , the short-term zero-crossing rate is relatively stable compared with other moments;

[0095] Analysis of time-frequency characteristics during the feeding period of the automatic feeding machine: The automatic feeding machine feeds large quantities of food between the two cages, and the sound signal at this moment is selected for time-frequency analysis to obtain the time-frequency diagram of this moment. During this period, there is aliasing of fish school sound signals. Due to the large-scale feeding of the automatic feeding machine, the fish school in the entire cage moves violently, and the fish school is most active. 2000-5000Hz is the frequency band where the fish school hits each other and splashes water, which has higher energy than manual feeding. When the audio signal of this period is played, the hissing and creaking friction sounds can be heard, as well as the sound of the fish swallowing accompanied by the bursting of bubbles on the water surface. Sound; In order to distinguish the two frequency bands of fish school signals that are aliased when the fish school is automatically fed on a large scale, a bandpass filter with a bandwidth of 600-2000Hz is used to filter the time domain signal, extract the time period when the golden pomfret noise appears, and calculate its short-time energy and short-time zero-crossing rate; a bandpass filter with a bandwidth of 200-7000Hz is used to filter the time domain signal, extract the time period when the croaker fish makes sounds, and calculate its short-time energy and short-time zero-crossing rate; the probability statistical distribution of the short-time energy and zero-crossing rate of the golden pomfret and the croaker fish is solved respectively, and the short-time characteristic differences of different fish schools are observed;

[0096] Analysis of time-frequency characteristics during the midnight period: At midnight, the activity of fish schools is low, and the frequency band of the noise they make is also low. There is only radiation noise from sporadic swimming and splashing of water between 2000-4000Hz, and the peak sound pressure level is about 90dB;

[0097] Comparison of vocal characteristics of different fish groups: Based on the above content, the vocalization and short-time characteristics of the three fish groups are compared. It can be seen that the vocalization mechanisms of the three fish groups are different, and their short-time energy and short-time zero-crossing rate characteristics are also different. The probability peak values ​​of short-time energy and short-time zero-crossing rate of the vocal signal of the caller fish are the largest. This is because the caller fish relies on the compression of the vocal muscles to drive the peritoneum to achieve vibration sound, which is mainly in the form of pulse trains and has high-order harmonics. The second is the golden pomfret, which makes sound by rubbing and slapping the tail and fins to produce a relatively wide-band signal form; grouper makes sounds by preying, chewing and swallowing food, with a relatively low frequency, and the probability peak values ​​of short-time energy and short-time zero-crossing rate are the smallest.

[0098] The present invention tests and analyzes the time-frequency domain characteristics of large-scale cage cultured fish within 24 hours, and selects the acoustic signals at three special times: manual feeding, automatic feeding machine feeding, and midnight for time-frequency analysis. First, the original signal is subjected to short-time Fourier transform to obtain the variation law of underwater noise frequency and signal intensity. According to the background noise of the ocean and the vocal characteristics of the fish, filters with different bandwidths are used to obtain fish acoustic signals of different frequency bands, and their short-time energy, short-time zero-crossing rate and power spectrum are calculated to compare and distinguish the vocal frequency bands, vocalization modes and short-time characteristics of the three fish. The results show that the spectral characteristics of fish can be determined by using signal processing technology. The vocalization frequency bands of different fish schools have different short-time characteristics, which can be used to distinguish the vocalization characteristics of different fish in the cage.

[0099] The present invention shows that: (1) feeding activities and external stimuli (such as the passing of fishing boats, power generation, etc.) will cause changes in the behavior and state of fish schools, and their acoustic characteristics can reflect their behavioral information. It can be seen from the power spectrum analysis of underwater acoustic signals that in the range of 100 to 5000 Hz, the energy of the acoustic signals emitted by fish schools is significantly higher than the background noise. The low-frequency background noise of the ocean is within 100 Hz, and the noise intensity gradually decreases and tends to be stable as the frequency increases after 5 kHz. It is mainly composed of the high-frequency background noise of the ocean.

[0100] (2) During the feeding stage of the fish in the cage, the frequency band of the sound emitted by the grouper school is mainly between 200 and 1000 Hz, the frequency band of the sound emitted by the golden pomfret is between 600 and 2000 Hz, and the frequency band of the noise generated by the fish slapping the water due to the competition for food is between 2000 and 5000 Hz. The sound signal of the croaker has a certain periodicity and high-frequency harmonics, with a cut-off frequency of about 7000 Hz. There is a certain frequency overlap in the sound of the golden pomfret and the croaker.

[0101] (3) The peak sound pressure level distribution of different fish schools is different. During the feeding stage, the peak sound pressure level of grouper is about 94dB, the peak sound pressure level of golden pomfret is about 108dB, and the peak sound pressure level of croaker is about 120dB.

[0102] (4) The frequency of the vocal signal of grouper is low, the short-term zero-crossing rate is low and the relative change is relatively stable; the vocal signal of golden pomfret is a non-harmonic pulse signal with a wide spectrum and a high probability distribution dispersion of the short-term zero-crossing rate; the vocal signal of croaker is in the form of a pulse train in the time domain and has the highest short-term energy and short-term zero-crossing rate.

[0103] The present invention is applied to the analysis of the correlation between the behavior of marine fish and the acoustic ecological environment by using passive acoustic technology; the present invention is reasonably designed and is worthy of being vigorously promoted.

[0104] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A statistical characteristic analysis method based on the acoustic signal characteristics of large cage cultured fish schools, characterized by: The statistical characteristic analysis method is as follows: Step 1: Data collection: Deploy hydrophones based on a large semi-submersible aquaculture platform; Step 2, test layout: In order to more clearly analyze the impact of external excitation on the fish school noise during the test process, track and record the events that may affect the noise signal during the data acquisition process; Step 3: Analyze the data: Use short-time analysis technology to study the characteristics of noise signals of different fish schools in the cage. Short-time energy describes the energy of each frame of sound signal. Its changing trend is consistent with the waveform of the sound signal, which can well reflect the time domain information of the sound. Assume that the time domain signal is x, add the Hanning window to process the frame and get the nth frame signal as x n (m), then x n (m) satisfies the following formula: x n (m)=w(m)x(n+m) 0≤m≤N-1 (1) The window function expression of the Hanning window used is: Where n = 0, 1T, 2T, ..., NT is the frame length, N is the window length; let the nth frame sound signal x n The short-time energy of (m) is expressed as E n The calculation formula is as follows: The short-time zero-crossing rate refers to the number of times the signal passes through the zero axis in each frame. The zero-crossing rate reflects the frequency information of the signal to a certain extent, and the expression is: Where sgn[] is the sign function; Since the sound is short-term stable in the time period of 10-30ms, the short-term analysis method takes 10-30ms as the frame length, and the length of each frame signal used for data processing is 25ms, and the frame shift is 10ms; Step 4. Results and analysis: Test signal time domain and frequency domain characteristics: In order to obtain the noise characteristics of the fish school, the data of the entire period is extracted every 5 minutes, the length of each data segment is 6326400, and the duration of each sample segment is 300s; The response characteristics of fish schools to events: During the artificial feeding period, at the feeding point of the grouper cage, the fish schools are more active and the noise intensity is higher than the background noise; Time-frequency characteristics analysis of artificial feeding: Artificial feeding was carried out in the grouper cage, and the sound signal at the current moment was selected for time-frequency analysis to obtain the short-time spectrum and power spectrum of the current period. In order to obtain the characteristics of the grouper group's feeding and vocalization stage, the noise frequency band of its vocalization was band-pass filtered with a filter bandwidth of 200-1000Hz, and its short-time energy and short-time zero-crossing rate were solved. It can be seen that the short-time energy of the signal during the grouper group's vocalization period is higher than that of other periods, and the peak energy is 8×10 -4 , the short-term zero-crossing rate is relatively stable compared with other moments; Analysis of time-frequency characteristics during the feeding period of the automatic feeding machine: The automatic feeding machine feeds large quantities of food between the two cages, and the sound signal at the current moment is selected for time-frequency analysis to obtain the time-frequency diagram of the current moment. During this period, there is an overlap of the fish school's sound signals. Due to the large-scale feeding of the automatic feeding machine, the fish school in the entire cage moves violently, and the fish school is most active. 2000-5000Hz is the frequency band where the fish school hits each other and splashes water, which has higher energy than manual feeding. When the audio signal of this period is played, the hissing and creaking friction sounds can be heard, as well as the bursting of bubbles on the water surface accompanied by the swallowing of fish. Sound; In order to distinguish the two frequency bands of fish signals that are overlapped when the fish are automatically fed on a large scale, a bandpass filter with a bandwidth of 600-2000Hz is used to filter the time domain signal, extract the time period when the golden pomfret noise appears, and calculate its short-time energy and short-time zero-crossing rate; a bandpass filter with a bandwidth of 200-7000Hz is used to filter the time domain signal, extract the time period when the croaker fish makes sounds, and calculate its short-time energy and short-time zero-crossing rate; the probability statistical distribution of the short-time energy and zero-crossing rate of the golden pomfret and the croaker fish is solved respectively, and the short-time characteristic differences of different fish schools are observed; Analysis of time-frequency characteristics during the midnight period: At midnight, the activity of fish schools is low, and the frequency band of the noise they make is also low. There is only radiation noise from sporadic fish swimming and splashing water between 2000-4000Hz, and the peak sound pressure level is 90dB; Comparison of vocal characteristics of different fish groups: Based on the above content, the vocalization and short-time characteristics of the three fish groups are compared. It can be seen that the vocalization mechanisms of the three fish groups are different, and the characteristics of their short-time energy and short-time zero-crossing rate are also different. The probability peak values ​​of short-time energy and short-time zero-crossing rate of the vocal signal of the caller fish are the largest; followed by the golden pomfret, which makes sounds by rubbing and slapping its tail and fins to produce a relatively wide-band signal form; grouper makes sounds by preying, chewing and swallowing food, with a relatively low frequency, and the probability peak values ​​of short-time energy and short-time zero-crossing rate are the smallest.

2. The statistical characteristic analysis method based on the acoustic signal characteristics of large-scale cage cultured fish schools according to claim 1 is characterized by: Requirements for deploying hydrophones as described in step 1: Place the hydrophones deep inside the cages outside the grouper culture area, golden pomfret culture area and automatic feed delivery area, with a water depth of 12.5m.

3. The statistical characteristic analysis method based on the acoustic signal characteristics of large-scale cage cultured fish schools according to claim 2 is characterized in that: The hydrophone is a self-contained hydrophone, the model of the hydrophone is TC4037, the sensitivity is -192dB, the amplification factor is 40dB, and during the test, the system sampling rate of the hydrophone is 105440Hz.

Images