Research On Adaptive Echo Cancellation Algorithm In Digital Hearing Aid

Digital hearing aids can help improve the hearing levels of people with hearing loss,and it has become an indispensable part of the lives of many hearing patients.But because of the small size of the hearing aid,the microphone and the speaker is very close,which led to the digital hearing aids easy to form acoustic feedback,echo phenomenon generated,serious there will be harsh howling signal which not only hurt the human ears,buy also easily damaged hearing aids.Therefore,digital hearing aids must include the processing of acoustic feedback.In this paper,the adaptive echo cancellation algorithm in digital hearing aids is studied,an improved algorithm of applying echo delay estimation to adaptive echo cancellation is proposed in order to achieve real-time echo cancellation,and frequency howls detection algorithm is proposed,and then make simulations.This paper introduces three key technologies in digital hearing aids,including voice enhancement,loudness compensation,and echo cancellation.And each of the technology were described in principle and formula derivation.The acoustic feedback cancellation of the digital hearing aids is mainly based on the adaptive echo cancellation algorithm,the method of the least mean square error(LMS)and the normalized minimum mean square error(NLMS)are described in detail.The experiments results with Matlab about LMS and NLMS show that NLMS adaptive filter has better adaptability.As the echo path is often not fixed,resulting in echo transmission time delay is uncertain,to solve this problem,it is proposed to apply the echo time delay estimation to the NLMS adaptive echo cancellation algorithm,to realize the real-time goal of solving the echo problem of digital hearing aids.Proposed Generalized Cross Correlation(GCC)and Cross Power Spectrum Phase(CSP)method which are based on the correlation coefficient,and adaptive Least Mean Square Error(LMS)Method which is based on path transfer function for echo delay estimation.The GCC method and the CSP method estimate the delay time by detecting the peak of the correlation coefficient.The simulation results show that the CSP effect is better than the GCC,and the adaptive echo cancellation algorithm with echo delay estimation is superior to the algorithm that does not do the echo delay estimation.The howling signal is a signal that is self-excited at a certain fixed frequency,and the howling detection in time and frequency domain is proposed for these features.In the frequency domain,calculate the peak frequency energy ratio,the peak point and adjacent point energy ratio,the peak duration,in the time domain calculate the correlation coefficient and the TKEO envelope,if the threshold is exceeded,judged that the signal is a howling signal.According to the howling test results,control the howling signal.The experimental show that the method can not completely avoid the occurrence of howling,only can reduce the probability of howling occurred.