Abstract
A speech enhancement method in the DCT domain using novel hard and soft thresholding criteria is proposed. Unlike other reported research results, the level of the trace of the signal present in the high frequency region of the DCT coefficients is estimated to obtain the critical threshold parameter, the bias compensated noise level. The results show that the proposed method is more effective for a wide range of SNRs.
Original language | English |
---|---|
Pages (from-to) | 669-670 |
Number of pages | 2 |
Journal | Electronics Letters |
Volume | 38 |
Issue number | 13 |
DOIs | |
Publication status | Published - 20 Jun 2002 |
Externally published | Yes |