Study And Hardware Implementation Of Microphone Array Speech Enhancement Algorithm In Indoor Environment

With the development of science and technology,speech signal has become an important medium of Human-computer interaction.Compared with the traditional single-channel microphone system,microphone array system can more effectively pick up target speech because of the characteristics of spatial directivity and high signal gain so that microphone array system is widely used in speech processing systems such as smart homes and wearable devices.In the indoor environment,the performance of the microphone array will decrease due to room reverberation and various noise,which will affect the human-computer interaction experience.To solve above problems,in this paper a new microphone array speech enhancement algorithm was proposed:using wavelet transform to optimize the traditional Generalized Sidelobe Canceller?GSC?speech enhancement algorithm and using convolutional neural network as post-filtering algorithm.The proposed algorithm will be implemented in FPGA for hardware acceleration and function verification.The main work of this paper is as follows:?1?In the indoor environment,the blocking matrix in the GSC cannot completely block the target speech signal,which causes speech leakage and affects the quality of enhanced speech.To solve this problem,a new GSC speech enhancement structure based on wavelet transform is proposed:taking place of blocking matrix with wavelet transform.Using wavelet decomposition and reconstruction to extract the noise signal of each microphone array channel to reduce speech leakage.?2?GSC speech enhancement algorithm can effectively suppress coherent noise in space,but there are still a lot of non-coherent and weak coherent noise in the enhanced speech signal.In order to solve this problem,convolutional neural network speech enhancement algorithm is used as post speech enhancement algorithm to improve the ability of the algorithm to suppress different types of noise.?3?The simulation of the proposed microphone array speech enhancement algorithm was performed.Compared with some existing algorithms in different noise environments,the results show that in the SNR?signal to noise ratio?range of-10d B-10d B,under the influence of Gaussian white noise and speech interrupt,the average score of speech perception quality evaluation?PESQ?of this method is increased by 24.48%on average;under the influence of Babble noise and speech interrupt,the average score of PESQ of this method is increased by17.44%;Under the influence of pink noise and speech interrupt,the average PESQ score of this method is increased by 21.09%on average;Under the influence of real indoor noise,the average PESQ score of this method is increased by 18.74%on average.?4?The hardware verification result of the proposed algorithm demonstrates that the output of the FPGA is nearly the same as the output of the software,with a maximum error of2^-11.The PESQ score is decreased by 2.65%,which proves the correctness of the function of hardware implementation within the error range.