An Improved MELP Speech Coder

This paper presents several strategies to improve the performance of MELP speech coder. A new approach is given to analyze and synthesize the transition segments which are called the transition frame. This can improve the accuracy of parameter estimation and reduce the noise caused by classing the transition frame simply into U/V frame. A new frame type decision algorithm with dynamic U/V threshold related to the pitch lag is introduced which classifies the speech frame into three types: voiced, unvoiced and transition. The classifier can reduce the U/V decision errors and avoid excessive switching between voiced frame and unvoiced frame. These strategies are incorporated into a speech codec M-MELP. The quality tests show that M-MELP is superior to the MELP coder, especially for high-frequency female speakers.