Abstract
A safe environment for drivers is very important for life today, but handset without hand-free system in a car always makes it dangerous to talk when driving. Thus an automatic speech recognition system is needed, but serious noise effect in car environment makes speech recognition result drastically decayed. A wavelet-based adaptive noise cancellation system is proposed to combat this problem and tries to increase the recognition rate. The MAT-160 database is used in the experiment for the training data, and the recognition result in the proposed system is six times better than the result without the proposed system when SNR is -5, and two times when SNR is 0.