A bird acoustic diversity index method with low sensitivity to noise

Abstract

This invention discloses a method for a bird acoustic diversity index with low sensitivity to noise. The method first obtains the time-frequency power spectrum of field bird call monitoring data during the analysis period using short-time Fourier transform. Then, based on narrowband noise level estimation, it calculates candidate floating detection thresholds for each frequency point that meet local signal-to-noise ratio requirements. The full-scale relative level threshold is used as the lower limit to obtain the frequency-varying detection threshold after binarization of the time-spectrum. Finally, the Shannon index is calculated by counting the number of time-frequency points in each sub-band where the power is higher than the detection threshold, resulting in the acoustic diversity index based on frequency-varying threshold detection. This invention effectively addresses the fundamental limitation of ADI (Ambient Audio Diversity Index) being highly sensitive to noise, while remaining fully compatible with the ecological interpretations and findings of existing ADI studies. It provides a new tool for achieving stable and reliable rapid assessment of bird species diversity, and is of great significance for ecological and environmental acoustic monitoring.

Description

Technical Field

[0001] This invention relates to the field of rapid assessment of bird species diversity and acoustic signal processing, specifically to a method for a bird acoustic diversity index that is low in sensitivity to noise. Background Technology

[0002] In the face of global ecological and environmental changes, biodiversity monitoring is an increasingly urgent task. The activity level of bird calls is an important indicator reflecting the biodiversity and ecological health of an ecological region, and rapid assessment of bird species diversity is of great significance for biodiversity monitoring and research.

[0003] With the popularization of acoustic monitoring technology and the development of soundscape ecology, acoustic indices, as a method for rapidly assessing bird species diversity, have been widely applied in ecological monitoring and environmental assessment. Acoustic indices are typically calculated from bird calls recorded in the field using recording equipment. However, the recorded data is not pure bird calls but a mixed signal containing various noise influences. Existing acoustic indices are extremely sensitive to noise, and the actual calculated results often do not match the expected ecological significance, making acoustic indices unsuitable for stable and reliable rapid assessment of bird species diversity. Chinese Invention Patent Publication No. CN109714689B discloses a method for obtaining noise-suppressed acoustic indices based on a differential microphone array. This invention constructs two directional beams using a differential microphone array and then employs spatial filtering technology to eliminate the adverse effects of fixed-directional mechanical noise on the acoustic index calculation. Chinese invention patent application number CN202111072342.0 discloses a method, device and storage medium for determining acoustic index. This invention eliminates common environmental noises such as wind, cicadas and streams through noise reduction processing, while compensating for distortion in the recorded bird song signal, thereby reducing the variation in acoustic index results caused by differences in the performance of recording equipment.

[0004] However, in practical applications, even in quiet ecological areas free from non-birdsong noise and under ideal conditions where signal distortion caused by recording equipment is negligible, inherent background noise inevitably exists in the recorded data. Even if this inherent background noise is weak, the signal-to-noise ratio (SNR) of the recorded birdsong is uncertain because the intensity of the actual recorded birdsong signal is uncontrollable. Existing acoustic indices are fundamentally flawed in their design, failing to consider the impact of noise. This causes the actual value of the acoustic indices to deviate further from the ideal value corresponding to a pure birdsong signal as the SNR decreases. Therefore, to achieve robust and reliable rapid assessment of bird species diversity, the design of existing acoustic indices must be fundamentally modified to obtain new acoustic index methods that are less sensitive to noise. However, the principles and methods for improving the robustness to noise generally differ for different acoustic indices. Summary of the Invention

[0005] The purpose of this invention is to provide a method for a bird acoustic diversity index that is less sensitive to noise.

[0006] The technical solution to achieve the purpose of this invention is: a method for a bird acoustic diversity index that is low in sensitivity to noise, comprising the following steps:

[0007] Step 1: Perform a short-time Fourier transform on the bird sound monitoring data collected during the analysis period to obtain its time-frequency power spectrum;

[0008] Step 2: Use the time-frequency power spectrum of the bird sound monitoring data obtained in Step 1 to estimate the average power of narrowband noise at each frequency point;

[0009] Step 3: Calculate the binarized detection threshold for each frequency point using the time-frequency power spectrum of the bird sound monitoring data obtained in Step 1 and the average power of the narrowband noise obtained in Step 2.

[0010] Step 4: Using the time-frequency power spectrum of the bird sound monitoring data obtained in Step 1 and the binarized detection threshold obtained in Step 3, perform 0-1 decision processing on the power of the time-frequency points of the bird sound monitoring data in each frequency point, and then count the number of time-frequency points in each sub-band whose power is higher than the detection threshold value of 1.

[0011] Step 5: Calculate the Shannon index using the number of frequency points with a value of 1 in each sub-band obtained in Step 4, and obtain the acoustic diversity index based on frequency-varying threshold detection.

[0012] In a second aspect, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method described in the first aspect.

[0013] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0014] Fourthly, the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in the first aspect.

[0015] Compared with existing technologies, the beneficial effects of this invention are as follows: 1) It corrects the design defects of existing indices in principle and proposes a new method for a bird acoustic diversity index that is less sensitive to noise, namely, an acoustic diversity index based on frequency-varying threshold detection; 2) While not affecting the ecological interpretation and discovery of existing acoustic index studies, it significantly suppresses the influence of changes in bird song SNR on the actual index calculation results and greatly reduces the lower limit of SNR to maintain index stability; 3) It adopts a completely passive acoustic monitoring method, which has no impact on the activities of wild birds; 4) The method of this invention is convenient to implement and easy to carry out, providing an efficient means for achieving stable and reliable rapid assessment of bird species diversity.

[0016] The present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0017] Figure 1 This is a flowchart of a method for an acoustic diversity index of birds that is low in sensitivity to noise.

[0018] Figure 2 These are time-domain waveforms and corresponding time-frequency spectra of monitoring recording data with the same bird call signal time-frequency distribution under different SNR conditions.

[0019] Figure 3 These are graphs showing the changes in the SNR of bird calls, calculated by FADI and ADI respectively for monitoring recording data with the same time-frequency distribution of bird call signals. Detailed Implementation

[0020] The literature (Villanueva-Rivera LJ, Pijanowski BC, Doucette J, et al. A primer of acoustic analysis for landscape ecologists[J].Landscape Ecology,2011,26(9):1233-1246.) discloses an acoustic diversity index (ADI), which is one of the commonly used indices, but its robustness to noise is low. This invention addresses the fundamental flaws of the widely used ADI in the literature "A primer of acoustic analysis for landscape ecologists" and proposes a method for a bird acoustic diversity index that is less sensitive to noise, namely, the frequency-dependent acoustic diversity index (FADI) based on frequency-varying threshold detection. The time-spectrum binarization processing of FADI employs narrowband floating detection thresholds set based on the average noise power at each frequency point while simultaneously meeting certain SNR requirements. The full-scale relative level (dB Full Scale, dBFS) threshold of ADI is used as the lower limit for each frequency point threshold, and the rest of the exponent calculation process is the same as ADI. The FADI proposed in this invention ensures that the exponent calculation results remain largely unchanged even when the SNR is below 0dB, provided that bird call signals have the same time-frequency distribution. Simultaneously, it maintains consistency with existing ADI values ​​at high SNR, thus not affecting the ecological interpretation and findings of existing ADI studies. This invention provides an efficient method for achieving stable and reliable rapid assessment of bird species diversity and offers a new approach for promoting the application of acoustic signal processing technology in ecological environment monitoring and assessment.

[0021] This invention discloses a method for a bird acoustic diversity index that is low insensitive to noise. This method is designed for ecological and environmental acoustic monitoring tasks. First, it obtains the time-frequency power spectrum of bird call monitoring data during the analysis period using short-time Fourier transform. Then, it calculates candidate floating detection thresholds for each frequency point that meet local signal-to-noise ratio requirements based on narrowband noise level estimation. The full-scale relative level (dBFS) threshold is used as the lower limit to obtain the frequency-dependent acoustic diversity index (FADI) after binarization of the time-spectrum. Finally, the Shannon index is calculated by counting the number of time-frequency points with power higher than the detection threshold in each sub-band, resulting in the FADI.

[0022] Combination Figure 1 The present invention provides a method for a bird acoustic diversity index that is low insensitive to noise, comprising the following steps:

[0023] Step 1: Perform a short-time Fourier transform on the bird sound monitoring data collected during the analysis period to obtain its time-frequency power spectrum:

[0024] Step 1-1: Perform frame processing on the bird sound monitoring data collected during the analysis period, with a frame shift of one frame length. Apply a Hanning window to each frame and then perform a Discrete Fourier Transform to obtain S(k,l), 1≤k≤K, 1≤l≤L, where k and l are the frequency point number and frame number, respectively, K is the number of frequency points within the analysis bandwidth, and L is the number of frames within the analysis period. In this invention, the analysis bandwidth is 0Hz-8kHz, the data sampling rate is 16kHz, the analysis period length is 600s, and the frame length is 10ms.

[0025] Step 1-2: Calculate the time-frequency power spectrum of the field bird sound monitoring data during the analysis period using the short-time Fourier transform S(k,l) obtained in Step 1-1.

[0026] P s (k,l)=|S(k,l)| 2 , 1≤k≤K, 1≤l≤L

[0027] Step 2: Estimate the average power of narrowband noise at each frequency point using the time-frequency power spectrum of the bird call monitoring data obtained in Step 1.

[0028] Step 2-1: Take all time-frequency points that do not contain bird song signals as noise time-frequency points, and denote the set of frame numbers of the noise time-frequency points corresponding to the k-th frequency point as ξ. k set ξ k The number of noise time-frequency points is M k ;

[0029] Step 2-2: Based on the set of noise time-frequency frame numbers ξ obtained in Step 2-1, k Calculate the average narrowband noise power at each frequency point.

[0030]

[0031] Step 3: Calculate the binarized detection threshold for each frequency point using the time-frequency power spectrum of the bird sound monitoring data obtained in Step 1 and the average power of the narrowband noise obtained in Step 2.

[0032] Step 3-1: Calculate the power value corresponding to the -50dBFS detection threshold of ADI using the time-frequency power spectrum of the bird sound monitoring data obtained in Step 1. Use this power value as the lower limit of the binarized detection threshold for each frequency point.

[0033]

[0034] In the formula, k L This invention uses k to represent the frequency index corresponding to the upper sideband of the low-frequency band where environmental noise energy is highly concentrated. L =20, corresponding to 200Hz;

[0035] Step 3-2: Using the narrowband noise average power obtained in Step 2, calculate the power value corresponding to the minimum signal-to-noise ratio γ1 required for bird sound detection at each frequency point, and use it as a candidate value for the binarization detection threshold at each frequency point.

[0036] η1(k)=γ1×Q n (k), 1≤k≤K

[0037] In this invention, γ1 = 4, meaning that the SNR of bird calls at the time-frequency point determined to be 1 must be greater than 4.

[0038] Step 3-3: Compare the candidate threshold values ​​for each frequency point obtained in Step 3-2 with the lower threshold values ​​obtained in Step 3-1, and take the maximum value of the two as the final binarized detection threshold for each frequency point.

[0039] η(k)=max{η1(k),η min}, 1≤k≤K

[0040] Step 4: Using the time-frequency power spectrum of the bird sound monitoring data obtained in Step 1 and the binarized detection threshold obtained in Step 3, perform 0-1 decision processing on the power of the time-frequency points of the bird sound monitoring data at each frequency point. Then, count the number of time-frequency points in each sub-band whose power is higher than the detection threshold (value 1).

[0041] Step 4-1: Perform 0-1 decision processing on the time-frequency power of the bird sound monitoring data at each frequency point to obtain the binarized time-frequency power spectrum of the bird sound monitoring data within the analysis period.

[0042]

[0043] Step 4-2: For the bird sound monitoring data obtained in Step 4-1, calculate the number of time-frequency points with a value of 1 within the analysis period using the binarized time-frequency power spectrum.

[0044]

[0045] In the formula, I represents the number of adjacent sub-bands within the analysis bandwidth. In this invention, I = 8, and each sub-band has the same bandwidth; θ i The set of frequency point indices within the i-th sub-band

[0046] θ i ={k|(i-1)B≤(k-1)Δf<iB},1≤i≤I

[0047] In the formula, B and Δf are the bandwidth of each sub-band and the bandwidth of each frequency point, respectively. In this invention, B = 1 kHz and Δf = 10 Hz are taken.

[0048] Step 5: Calculate the Shannon index using the number of frequency points with a value of 1 within each sub-band obtained in Step 4, to obtain the acoustic diversity index based on frequency-varying threshold detection.

[0049] Step 5-1: Normalize the number of frequency points with a value of 1 in each sub-band obtained in Step 4 to obtain the distribution ratio of the number of frequency points with a value of 1 in each sub-band.

[0050]

[0051] Step 5-2: Calculate the Shannon index for the distribution ratio of the number of frequency points in each sub-band when the value of 1 is obtained in Step 5-1, to obtain the FADI value.

[0052]

[0053] Figure 2 In the middle (a) and (b), the time-domain waveforms and corresponding time-frequency spectra of the monitoring recording data with the same time-frequency distribution of bird call signals are shown respectively when the SNR is 5dB and 35dB. Figure 3 The graphs showing the variation of the exponent values ​​calculated by FADI and ADI for monitoring recordings with the same time-frequency distribution of bird calls as a function of the sound frequency response (SNR) are presented. The analysis data duration for calculating the exponent under each SNR condition is 10 minutes. Figure 3 It can be seen that when bird call signals have the same time-frequency distribution, the FADI calculation results do not change significantly even when the SNR is below 0 dB, while the ADI calculation results show a clear monotonically decreasing change as the SNR decreases. This indicates that FADI has effectively solved the fundamental defects of ADI, thereby reducing the sensitivity of actual calculation results to noise. At the same time, the FADI value remains consistent with ADI at high SNR, and will not affect the ecological interpretation and findings of existing ADI studies.

[0054] This invention is easy to implement and can effectively solve the problem that existing acoustic indices are easily affected by noise. When bird song signals have the same time-frequency distribution, it ensures that the actual calculated value of FADI does not change significantly when the signal-to-noise ratio of bird song drops to 0dB. At the same time, it can maintain the same value as the existing ADI at high signal-to-noise ratios. Thus, it can effectively solve the fundamental defect that ADI is very sensitive to noise, while also being fully compatible with the ecological interpretation and findings of existing ADI studies. It provides a new tool for achieving stable and reliable rapid assessment of bird species diversity and is of great significance for acoustic monitoring of the ecological environment.

[0055] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for a bird acoustic diversity index with low sensitivity to noise, characterized in that, The steps are as follows: Step 1: Perform a short-time Fourier transform on the bird call monitoring data during the analysis period to obtain its time-frequency power spectrum, as follows: Step 1-1: Perform frame processing on the bird sound monitoring data collected during the analysis period, with a frame shift of one frame length; after adding a Hanning window to each frame of data, perform discrete Fourier transform processing to obtain S(k,l), 1≤k≤K, 1≤l≤L, where k and l are the frequency point number and frame number, respectively, K is the number of frequency points within the analysis bandwidth, and L is the number of frames within the analysis period; Step 1-2: Calculate the time-frequency power spectrum of the field bird sound monitoring data during the analysis period using the short-time Fourier transform S(k,l) obtained in Step 1-1. P s (k,l)=|S(k,l)| 2 ,1≤k≤K,1≤l≤L Step 2: Using the time-frequency power spectrum of the bird call monitoring data obtained in Step 1, estimate the average power of narrowband noise at each frequency point, as follows: Step 2-1: Take all time-frequency points that do not contain bird song signals as noise time-frequency points, and denote the set of frame numbers of the noise time-frequency points corresponding to the k-th frequency point as ξ. k set ξ k The number of noise time-frequency points is M k ; Step 2-2: Based on the set of noise time-frequency frame numbers ξ obtained in Step 2-1, k Calculate the average narrowband noise power at each frequency point. Step 3: Using the time-frequency power spectrum of the bird sound monitoring data obtained in Step 1 and the average power of the narrowband noise obtained in Step 2, calculate the binarized detection threshold for each frequency point, as follows: Step 3-1: Calculate the power value corresponding to the -50dBFS detection threshold of ADI using the time-frequency power spectrum of the bird sound monitoring data obtained in Step 1. Use this power value as the lower limit of the binarized detection threshold for each frequency point. In the formula, k L The frequency number corresponding to the upper sideband of the low-frequency band where environmental noise energy is highly concentrated; Step 3-2: Using the narrowband noise average power obtained in Step 2, calculate the power value corresponding to the minimum signal-to-noise ratio γ1 required for bird sound detection at each frequency point, and use it as a candidate value for the binarization detection threshold at each frequency point. η1(k)=γ1×Q n (k),1≤k≤K Step 3-3: Compare the candidate threshold values ​​for each frequency point obtained in Step 3-2 with the lower threshold values ​​obtained in Step 3-1, and take the maximum value of the two as the final binarized detection threshold for each frequency point. η(k)=max{η1(k),η min },1≤k≤K Step 4: Using the time-frequency power spectrum of the bird sound monitoring data obtained in Step 1 and the binarized detection threshold obtained in Step 3, perform 0-1 decision processing on the power of the time-frequency points of the bird sound monitoring data in each frequency point, and then count the number of time-frequency points in each sub-band whose power is higher than the detection threshold value of 1. Step 5: Calculate the Shannon index using the number of frequency points with a value of 1 in each sub-band obtained in Step 4, and obtain the acoustic diversity index based on frequency-varying threshold detection.

2. The method for a bird acoustic diversity index with low sensitivity to noise as described in claim 1, characterized in that, k L =20, corresponding to 200Hz.

3. The method for a bird acoustic diversity index with low sensitivity to noise as described in claim 1, characterized in that, Step 4 uses the time-frequency power spectrum of the bird call monitoring data obtained in Step 1 and the binarized detection threshold obtained in Step 3 to perform 0-1 decision processing on the power of the time-frequency points of the bird call monitoring data at each frequency point. Then, the number of time-frequency points with power values ​​higher than the detection threshold (value 1) in each sub-frequency band is counted, as follows: Step 4-1: Perform 0-1 decision processing on the time-frequency power of the bird sound monitoring data at each frequency point to obtain the binarized time-frequency power spectrum of the bird sound monitoring data within the analysis period. Step 4-2: For the bird sound monitoring data obtained in Step 4-1, calculate the number of time-frequency points with a value of 1 within the analysis period using the binarized time-frequency power spectrum. In the formula, I represents the number of adjacent sub-bands within the analysis bandwidth, and θ i θ is the set of frequency indexes within the i-th sub-band. i ={k|(i-1)B≤(k-1)Δf<iB},1≤i≤I In the formula, B and Δf are the bandwidth of each sub-band and the bandwidth of each frequency point, respectively.

4. The method for a bird acoustic diversity index with low sensitivity to noise as described in claim 3, characterized in that, Step 5 uses the number of frequency points with a value of 1 in each sub-band obtained in Step 4 to calculate the Shannon index, obtaining the Acoustic Diversity Index (FADI) based on frequency-varying threshold detection, as follows: Step 5-1: Normalize the number of frequency points with a value of 1 in each sub-band obtained in Step 4 to obtain the distribution ratio of the number of frequency points with a value of 1 in each sub-band. Step 5-2: Calculate the Shannon index for the distribution ratio of the number of frequency points in each sub-band when the value of 1 is obtained in Step 5-1, to obtain the FADI value.

5. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method described in any one of claims 1-4.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method described in any of claims 1-4.

7. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the method described in any one of claims 1-4.

Images