Design of echo cancellation and noise elimination for speech enhancement

In this paper, we present the new methods to cancel echo especially for the short path (e.g. car) and eliminate noise for speech enhancement. An adaptive filter based on the delayed error least mean square algorithm is used to cancel echo. The external noise is eliminated and the clean speech is estimated by using the Kalman filter and the spectral subtraction technique. To be suitable for use in consumer electronics, we also design a high speed and flexible VLSI architecture of the adaptive filter for canceling echo. The architecture has hardware utilization efficiency of 100%, and we can easily scale the filter without reducing the throughput rate. The 0.6 (mu)m CMOS SPTM standard cells technology has been used to implement the chip. The experimental results demonstrate the effectiveness of the proposed methods.