Abstract :
The speech enhancement is one of the effective techniques to solve speech degraded by noise. In this paper a fast speech enhancement method for noisy speech signals is presented, which is based on improved Kalman filtering. The conventional Kalman filter algorithm for speech enhancement needs to calculate the parameters of AR (auto-regressive) model, and perform a lot of matrix operations, which usually is non-adaptive. The speech enhancement algorithm proposed in this paper eliminates the matrix operations and reduces the calculating time by only constantly updating the first value of state vector X(n). We design a coefficient factor for adaptive filtering, to automatically amend the estimation of environmental noise by the observation data. Simulation results show that the fast adaptive algorithm using Kalman filtering is effective for speech enhancement.
In the past several years, there were many applications in speech enhancement based on Kalman filtering algorithm. Those methods were proposed by [1]-[7]. Most of those methods need to estimate the parameters of AR model at first, and then perform the noise suppression using Kalman filtering algorithm. In this process, the calculations of LPC (linear prediction coding) coefficient and inverse matrix greatly increase the computational complexity of the filtering algorithm. Although these methods can achieve a good filtering efficiency, the noise suppressed signal may deteriorate the quality of the speech signal dependent on estimation accuracy of the parameters of the AR model. [2] and [3] have been given a simple Kalman filtering algorithm without calculating LPC coefficient in the AR model, but the algorithm still contains a large number of redundant data and matrix inverse operations. In addition, the algorithm is non-adaptive. Simulation results show that, compared with the conventional Kalman filtering algorithm, the fast adaptive algorithm of Kalman filtering is more effective. At the same time, it reduced its’ running time without sacrificing quality of the speech signal. It also has good adaptability to improve the algorithm robustness