Acoustic Sensor Array Data Acquisition And Processing System Research Based On FPGA

The acoustic signal processing system modeled by acoustic array has been widely applied to various fields,such as smart conference,smart home,smart driving and diverse smart devices.“Directional Sound-Sensing” is a multi-channel speech enhancement technique which uses acoustic array as the implementation carrier,broadband beamforming as the implementation algorithm,and smart devices as the application target,aiming at providing high-quality audio input for the terminal smart applications.The system based on Field Programmable Gate Array(FPGA)can meet the technical requirements of complex computing,precise control,stable interaction,high throughput and low latency of acoustic-array oriented embedded systems,and is flexibly applied in different acoustic array systems.The goal of this research is to design an FPGA-based acoustic array system using ZYNQ-So C to achieve the function of Directional Sound-Sensing and apply it to different acoustic environments in real time to complete the established performance tests.The main innovative work of this work is that: In order to solve the key technical difficulties of “Directional Sound-Sensing”,including the “algorithm-level” breakthrough: short sound-capture distance,poor suppression of non-stationary speech interferers and inaccurate estimation of direction of arrival in complex acoustic environments,as well as the “system-level” breakthrough: low throughput,weak real-time,poor compactness,and high power consumption due to conventional DAQ+DSP/GPU/x86 Hardware schemes,two on-chip FPGA-based integrated acquisition-processing-transmission schemes for acoustic array systems and on-chip Directional Sound-Sensing processing algorithms implemented to each acoustic array system are designed.The proposed systems optimize the performance of Directional Sound-Sensing and show potential for applications in Directional Sound-Sensing and terminal intelligent voice interaction in complex environments.The research includes:1.Hardware system design: To address the above limitations of existing systemlevel hardware solution,design two sets of acoustic array systems based on FPGA,a 48-element microelectromechanical system(MEMS)digital microphone array system and a4-element fiber optic distributed acoustic sensing array system,respectively.This research will implement FPGA-based digital customization and on-chip design of acquisition-processing-transmission.Also,complete the hardware circuit design of the corresponding system,including PCB fabrication,peripheral configuration,and prototype assembly.2.Hardware accelerated algorithm design: To address the above performance limitations of the existing algorithm level,design FPGA-based hardware pipelined acceleration design for the two types of acoustic array systems respectively to achieve high-quality Directional Sound-Sensing.For the digital MEMS microphone array system,the hardware acceleration focuses on on-chip design of acquisition-processingtransmission and Audio-Visual-based broadband acoustic beamforming algorithm;for the fiber optic distributed acoustic array system,the hardware acceleration focuses on,in addition to the above,finite-point periodic trigger acquisition based on high-speed ADC,Trace stream to temporal series transposition operation,and 3×3 coupler differentiate and cross-multiply demodulation algorithm based on time division multiplexing optimization.3.Performance test experimentations: For the digital MEMS microphone array system,it satisfies real-time and stable operation,as well as achieves segmental SINR improvement of 12.7 d B,extensive short-time objective intelligibility(ESTOI)of 0.96,sound-capture distance of 15 m,and beam angle resolution within 20° in the given test environment.For the fiber optic distributed acoustic array system,it satisfies real-time and stable operation with an RMSE of 2.4° in sound source localization,4.3 d B for segmental SINR improvement,and 0.92 for ESTOI.