Research On Noise Reduction Using Microphone Array And Application In Passenger Information System

Speech noise reduction and speech enhancement are important research contents in the field of speech signal processing,and noise reduction based on microphone array has become a hot point in study.The normal operation of all trades and professions is inseparable from speech communication,and as the quality of speech signals determines whether the communication can proceed normally,noise reduction and speech enhancement are playing more vital roles in communication.With the development of public transportation,the demand for speech communication in the passenger information system is also increasing,and the application of noise reduction based on microphone array has become the current trend of development.At present,the passenger information system mostly relies on a single microphone to collect audio,and lacks noise reduction processing for speech signals.Therefore,it is expected to design a noise reduction software to provide speech enhancement for the passenger information system,and at the same time,the single microphone is replaced by microphone array,and the effect of noise reduction of the software is improved during environment of strong diffuse noise and complex noise.In this thesis,noise reduction based on microphone array is studied,and the adaptive noise canceller of generalized sidelobe canceller is improved.The signal-to-noise ratio of output signal to reference noise signal and proportion of noise in output signal are used to control the updating of weight coefficient of the adaptive noise canceller.Compared with the original algorithm,whether the output signal is the desired can be judged more accurately by the proportion of noise,thus it can avoid the misjudgment caused by excessive power spectral density of diffuse noise.At the same time,aiming at the large steady-state error of the normalized least mean square algorithm,the variable step-size least mean square algorithm is improved,and a new step-size updating formula is given,which effectively reduces the noise component in the output signal.When detecting the activity of speech signal in multichannel postfiltering algorithm,it is necessary to estimate the power spectral density of noise signal.In order to improve the accuracy of noise estimation,this thesis introduces the change rate of power spectral density of noise signal,which can capture the changing of noise signal more accurately.It will allow the estimated noise signal adjust quickly by controlling noise estimation with the change rate of power spectral density of noise signal,so as to detect the desired speech more accurately and improve the robustness of the algorithm.Finally,the feasibility of the proposed algorithm is verified by experiment compared with the original algorithm.The experiment result shows that the proposed algorithm can effectively reduce the noise component in the output signal,and it can still ensure the effect of speech enhancement in the environment under strong diffuse noise.