Research On Multi-channel Audio Signal Acquisition And Analysis System

Since multi-channel audio signal acquisition and processing technology is widely used in daily life,construction sites,military,aerospace and other fields,sound source localization and sound classification methods have become one of the hot research issues.The traditional audio signal acquisition system has problems such as large equipment volume,insufficient real-time processing capability,and poor sampling synchronization.In this paper,the AM5708 multi-core microprocessor is used as the core chip,the system hardware design is completed,and the dual operating system framework of Linux-RT+TI-RTOS is built,which lays the foundation for the subsequent algorithm implementation.Aiming at the problem of sound source localization,this paper deduces a spherical coordinate system localization model based on the topology of the five-element microphone,and obtains three spherical coordinates of the sound source.For the synthesized five sound signal time series,the cross-correlation delay estimation algorithm is used to extract four delay estimates as the input of the positioning model,and different forms of perturbation methods are applied through MATLAB to verify the correctness and robustness of the positioning model and the delay estimation algorithm.In this paper,a large number of simulation experiments are carried out to study the influence of sampling frequency on the accuracy of sound source localization distance factor,and then the appropriate sampling frequency was determined,which can not only meet the localization requirements,but also be easy to implement on the embedded system.The dimension of the feature vector and the parameter setting of the classifier directly affect the accuracy of sound classification.After many simulation tests,the kernel function parameters of the SVM classifier and the dimension of the feature vector are finally determined,and 13-dimensional feature vectors were respectively constructed for the four periodic sound signals,which are used as the input of the SVM classifier to achieve accurate sound classification.Based on the positioning model and the trained SVM classifier used in this paper,a C language application for dual-operating systems of ARM+DSP multi-core heterogeneous embedded hardware platform is written.The ARM side is responsible for system management tasks for interface configuration and serial communication.The DSP side is responsible for the synchronous acquisition and storage of five channels of audio signals,as well as tasks with high real-time requirements such as FFT operations,delay estimation,positioning algorithms,and SVM classification.Task.real-time data exchange is performed between ARM and DSP through IPC.The laboratory test results show that the multi-channel audio signal acquisition and analysis system designed in this paper has stable performance,the positioning error is less than 4%,and the classification accuracy reaches 95.25%.The results verify the feasibility and accuracy of the designed system.