Research On Echo Cancellation Algorithm For Digital Hearing Aids

With the advent of an aging society,the number of elderly patients with hearing loss is increasing year by year.And the hearing rehabilitation of patients with hearing loss will face serious challenges.For patients with mild,moderate or even moderate to severe hearing loss,wearing digital hearing aids is the most effective and convenient method at present.As one of the key algorithms of digital hearing aids,the echo cancellation algorithm seriously affects the hearing experience of patients wearing hearing aids.The research work of this thesis mainly focuses on the echo cancellation algorithm in digital hearing aids.Based on the theory of classical echo cancellation algorithms,the adaptive filter based on proportional coefficient and the parametric Wiener based on spectrum correction are deeply studied.Besides,the improvement schemes are proposed respectively.Finally,the methods of theoretical analysis and software simulation are used to verify the effectiveness of the proposed strategies.The main research and innovation work of the thesis are as follows.(1)The principle and characteristic of speech signal,and the model of human hearing system are studied.The hardware structure of digital hearing aids is analyzed,and the mainstream software algorithms are expounded.Two types of common echo cancellation algorithms in digital hearing aids are mainly studied,including the adaptive filtering algorithm based on time/frequency domain and the scheme of spectrum correction in the field of speech enhancement.Then the evaluation indicators are summarized.(2)A L0-IPNLMS adaptive algorithm based on set membership filtering is proposed.Firstly,the coefficient proportional NLMS and its improved algorithms are studied.Secondly,considering the characteristics of low power consumption of digital hearing aids,the L0-IPNLMS algorithm based on the SMF theory is proposed with the data selection update characteristics.By establishing the connection between the error signal and the step size control,not only the strategy of variable step size is realized,but also the calculation of many redundant data is discarded.In addition,through the dynamic analysis of the step size of the improved algorithm,it is verified that the redundant update calculation is correctly discarded during the convergence process of the algorithm.It saves a certain amount of computing resources.Finally,the improved algorithm is compared with a variety of commonly algorithms.And it is verified that the improved algorithm not only has good echo cancellation performance,but also greatly robust for low SNR.(3)An improved parametric Wiener filter algorithm based on optimal smoothing factor is proposed.The working principle of echo canceller based on the basic Wiener filter and the prior signal-to-echo ratio are studied,respectively.And the echo suppression model based on the parametric Wiener filter is constructed.In this model,the acoustic echo spectrum is firstly estimated by using the coloring effect.Then,according to the inhibition curve of the parametric Wiener filter,a parameter adaptive strategy based on the posterior signal-to-echo ratio is designed.In addition,the cost function is established according to the minimum-mean square estimation criterion.What’s more,the priori SNR estimation algorithm based on the optimal smoothing factor is deduced.Finally,the effectiveness of the optimal smoothing factor and parameter adaptive strategy in the improved algorithm is verified by theoretical analysis and simulation experiments.