Design And Implementation Of Classroom Sound Pickup System Based On Microphone Array

Under the background of the rapid popularization of education informationization,many schools have introduced many teaching auxiliary equipment,such as recording classroom,teaching machine and so on.However,in the rich auxiliary teaching tools,teachers often neglect the most basic needs of amplification.In the traditional teaching process,teachers usually wear audio pickup devices such as ear-hanging microphones or hand-held microphones.The microphone picks up voice signals and transmits them to speakers in wired or wireless ways to complete sound amplification,which has problems such as uncomfortable wearing of microphones.In recent years,classroom sound amplification systems have begun to use hoisting microphones instead of traditional wearable microphones.The hoisting microphone can pick up the sound in a fixed area by lifting a single directional microphone in front of the platform,so as to truly meet the basic needs of "insensitive" sound amplification for teachers.However,there are some problems in this way,such as too single picking up area.Based on the signal processing technology of microphone array,this thesis presents a design method of ring microphone array as sound pickup device.The process of far field speech picking is divided into three stages: endpoint detection,speaker source location and speech enhancement in specified direction.The main work of this thesis is as follows:1.Compare the performance of two speech endpoint detection algorithms,the double-threshold method and the subband spectral entropy method.The double threshold method is difficult to determine the threshold value,and the detection results are unstable under different SNR conditions.The result of subband spectral entropy method is more accurate,and under the same threshold value,different SNR and different speech detection results are basically the same,which has high stability and universality.2.Compare the performance of two sound source localization algorithms,DAS and MVDR.MVDR algorithm is better than DAS algorithm in both accuracy and average deviation Angle.In the anti-reverberation performance comparison experiment,MVDR algorithm is stronger than DAS algorithm in suppressing the beam sidelobes with the increase of frequency,and the anti-reverberation ability of MVDR algorithm is significantly higher than DAS algorithm when the frequency is below 2KHz.3.Compare the performance of delayed sum beamforming algorithm and generalized Sidelobe cancellation(GSC)beamforming algorithm.Under different frame lengths,the output SNR and expected signal distortion of DAS and GSC algorithms will increase with the increase of frame lengths.Under different input SNR conditions,the output SNR performance of GSC algorithm is better than that of DAS algorithm,and compared with DAS algorithm,GSC algorithm can effectively suppress signals in other incoming directions by using array redundancy.4.Based on the classroom environment,this thesis uses Re Speaker Mic Array as the microphone Array hardware device,transplants the algorithm and runs it in Raspberry Pi Linux C++ environment,and builds a classroom sound pickup system based on microphone Array.The validity of the speech endpoint detection algorithm,the sound source localization algorithm and the beamforming speech enhancement algorithm are tested.The feasibility of the system is verified by experiments.