An Non-Contact Speech Enhancement Algorithm Based on Lifting Scheme

The traditional method for detecting speech signal needs the help of microphone, which must be placed closely to the body of human beings. To some extent, this method would bring several inconveniences. Non-contact speech detection method breaks through the limitation of the traditional method, this new kind of speech obtaining method can detect speech signal quite well even in strong noisy background. However, this non-contact speech detecting system also produces some electromagnetic noise and circuit noise, which reduced the quality of radar speech signal. Therefore, based on the good time-frequency analyze performance, the lifting scheme was also proposed in this paper to remove noise from radar speech. Comparing to classical enhancement algorithm, such as spectral subtraction and Wiener filter, the proposed algorithm can remove the component of noise availably and reserve the original pure speech signal in a promising way.