CN-121978625-A - Intelligent acoustic positioning system for sick pigs in pig house

CN121978625ACN 121978625 ACN121978625 ACN 121978625ACN-121978625-A

Abstract

An intelligent acoustic positioning system for a sick pig in a pig house relates to the field of sound recognition and positioning. The problems of poor voice recognition accuracy and large positioning error of the existing acoustic positioning method for the sick pigs in a complex acoustic environment are solved. The invention adopts the positioning principle of complementation of time difference and phase difference, combines the self-adaptive beam forming technology, solves the problem of inaccurate sound source positioning in the pig house environment with strong reflection and high noise, and carries out multi-feature extraction and intelligent diagnosis on the enhanced sound data through the sick pig sound recognition unit, thereby improving the recognition accuracy. The method is mainly used for positioning diseased pigs in a complex cultivation environment.

Inventors

YANG QIHAN
YAO YUCHANG
JIANG ZHIHAN
LIU GUANSONG
LIU HENGYAN
SU CHANG

Assignees

东北农业大学

Dates

Publication Date: 20260505
Application Date: 20260123

Claims (10)

1. An intelligent acoustic positioning system for a sick pig in a pig house, comprising: The collecting unit is used for collecting mixed sounds in the pigsty through the microphone array; the sound source positioning and noise reduction unit is used for calculating the arrival time difference and the phase difference of each target sound source reaching any two microphones according to the mixed sound collected by each microphone in the array, and reversely deducing the direction angle of the target sound source by using a direction-of-arrival estimation algorithm according to all the arrival time difference, the phase difference and the geometric shape of the microphone array; And the sick pig sound identification unit is used for carrying out multi-feature extraction and intelligent diagnosis on the enhanced sound data and judging whether the sound emitted by the target sound source corresponding to the enhanced sound data is from sick pigs.
2. An intelligent acoustic positioning system for pig house diseases according to claim 1, wherein the microphone array is a specific geometry array of 6 microphones.
3. An intelligent acoustic positioning system for a pig house according to claim 1, wherein the specific geometry is circular or square.
4. The intelligent acoustic positioning system for pig house sick pigs according to claim 1, wherein the calculation of the arrival time difference and the phase difference of each target sound source to any two microphones is realized by a cross-correlation function.
5. An intelligent acoustic positioning system for a pig house according to claim 1, wherein the pig house sound recognition unit comprises: The characteristic extraction module is used for extracting a Mel frequency cepstrum coefficient and a linear prediction coding coefficient from the enhanced sound data as characteristic parameters, wherein the Mel frequency cepstrum coefficient is used as a tone fingerprint characteristic; And the intelligent diagnosis module predicts the enhanced sound data input into the intelligent diagnosis module and the Mel frequency cepstrum coefficient and the linear predictive coding coefficient corresponding to the enhanced sound data by adopting a three-channel convolutional neural network model after pre-training to obtain a diagnosis result, wherein the diagnosis result is a sick pig or a non-sick pig.
6. The intelligent acoustic positioning system for pig house diseases according to claim 5, wherein the implementation manner of extracting mel frequency cepstral coefficients from the enhanced sound data is as follows: the enhanced sound data is converted into a frequency spectrum through Fourier transformation, the Mel frequency spectrum energy in the frequency spectrum is extracted through a Mel filter bank, the logarithm of the Mel frequency spectrum energy is measured, and the Mel frequency cepstrum coefficient is obtained through discrete cosine transformation.
7. The intelligent acoustic positioning system for pig house disease according to claim 5, wherein the linear prediction coding coefficient is extracted from the enhanced sound data by the following implementation manner: pre-emphasis processing is carried out on the enhanced sound data; Frame-dividing the pre-emphasized enhanced sound data into frame sequences with frame length of 20-30 ms and frame overlapping rate of 50%; based on the front in each frame signal after application of the hamming window The audio signal sampling value at each moment is used for constructing a linear prediction equation; The method aims at the minimum mean square value of the prediction error, solves the linear prediction equation by an autocorrelation method to obtain a group of prediction coding coefficients, wherein the group of prediction coding coefficients comprises And predictive coding coefficients.
8. The intelligent acoustic positioning system for pig house disease according to claim 7, wherein the expression of the linear prediction equation is: Wherein, the method comprises the steps of, For the current time in each frame signal after application of the hamming window Is provided with a plurality of audio signal samples, ; Is the first Linear predictive coding coefficients; is the prediction order; is a prediction error; Is in front of Audio signal samples at individual moments in time.
9. The intelligent acoustic positioning system for pig house diseases according to claim 1, wherein the intelligent diagnosis module comprises an input layer, a flattening layer, a first full-connection layer, a second full-connection layer and an output layer which are sequentially arranged in the propagation direction; The input layer is used for fusing the received three-channel data into a three-channel fusion feature map by adopting a feature splicing and attention mechanism fusion strategy; A flattening layer for converting the three-way fusion feature map into a one-dimensional vector; The first full-connection layer is used for carrying out global information integration and nonlinear combination learning on the one-dimensional vectors and outputting first feature vectors; the first full-connection layer is used for carrying out global information integration on the first feature vector and outputting a second feature vector; and the output layer is used for carrying out linear transformation and activation classification on the second feature vector, outputting the probability of each class, and taking the class corresponding to the maximum probability as a diagnosis result.
10. The intelligent acoustic positioning system for sick pigs in a pig house according to claim 1, further comprising a management platform for receiving and processing output results of the sound source positioning and noise reduction unit and the sick pig sound recognition unit for visual display and alarm.

Description

Intelligent acoustic positioning system for sick pigs in pig house Technical Field The present invention relates to the field of voice recognition and localization. Background With the large-scale and intensive development of pig breeding, epidemic disease prevention and control become the core problem restricting industrial benefit. Traditional pig detection relies on manual inspection, and is low in efficiency, high in omission factor and easy to cause cross infection, and intelligent technology is needed to improve epidemic disease monitoring capability. In the field of animal monitoring in farms, the prior art mainly comprises machine vision monitoring, wireless Sensor Network (WSN) monitoring, acoustic monitoring and the like. The acoustic monitoring technology has the unique advantages of non-contact, strong real-time performance, no limitation of illumination conditions, wide monitoring range, no dead angle basically and the like, and has great application potential in animal health and behavior analysis. Particularly, by combining Artificial Intelligence (AI) and internet of things (IoT) technologies and continuously collecting and analyzing sound signals of live pigs, the method is expected to become a key technical path for early identification and early warning of sick pigs. However, in a complex actual cultivation environment, a simple voice recognition technology is not enough to meet the requirement of accurate prevention and control, so that in an automatic cultivation process, the health condition of the live pigs is judged by utilizing an automatic technology through sound signals of the live pigs, and the position of sounding pigs is locked quickly, so that the method has practical significance. The main problems of the existing acoustic positioning technology for the sick pigs are summarized as follows: the positioning accuracy and reliability are insufficient, in complex acoustic environments such as a pig house with strong reflection and multiple noises (such as fan sounds, other pig sounds and equipment operation sounds), the existing sound source positioning technology (such as technology based on a single or simple array) is easily affected by reverberation and interference, the sound source direction is difficult to estimate stably and accurately, and the positioning error is large. The system has weak anti-interference capability, the pigsty background noise is various and has variable intensity, the existing system often lacks an effective self-adaptive noise reduction mechanism, so that the target diseased pig sound signal is submerged, the signal-to-noise ratio (SNR) is too low, and the accuracy of subsequent identification and positioning is seriously affected. The pertinence and the robustness of feature extraction are to be improved, namely, feature extraction of pig pathological sounds (such as cough and wheezing) is usually carried out directly along with general voice features (such as MFCC), and depth modeling of pig sounding physiological mechanisms and acoustic characteristic changes caused by diseases is lacked, so that the distinguishing degree and the robustness of the features in a complex environment are insufficient, and the model sound recognition accuracy is poor. In view of the above, the prior art has not been able to effectively solve the comprehensive problem of high-precision, high-reliability and high-real-time individuation positioning and recognition of the sound of the sick pigs in a complex acoustic environment, and the above problems need to be solved. Disclosure of Invention The invention aims to solve the problems of poor voice recognition accuracy and large positioning error of the existing acoustic positioning method for the sick pigs in a complex acoustic environment, and provides an intelligent acoustic positioning system for the sick pigs in a pig house. An intelligent acoustic positioning system for a pig house sick pig, comprising: The system comprises a collection unit, a sound source positioning and noise reduction unit, a sound source data acquisition unit and a sound source data acquisition unit, wherein the collection unit is used for collecting mixed sounds in a pig house through a microphone array, the sound source positioning and noise reduction unit is used for calculating arrival time differences and phase differences of each target sound source to any two microphones according to the mixed sounds collected by each microphone in the array, and reversely deducing a direction angle of the target sound source by utilizing a direction of arrival estimation algorithm according to all the arrival time differences, the phase differences and the geometric shapes of the microphone array; And the sick pig sound identification unit is used for carrying out multi-feature extraction and intelligent diagnosis on the enhanced sound data and judging whether the sound emitted by the target sound source corresponding to the enhanced sound data is fro