An improved 1.2 kb/s speech coder based on MELP

This paper presents key algorithms of an unproved 1.2 kb/s speech coder based on MELP. At 1.2 kb/s the MELP parameters for three consecutive frames are grouped into a super-frame and vectors are jointly quantized to obtain high coding efficiency. Besides fractional pitch and pitch doubling check, a sliding pitch analysis window method is proposed in this paper to improve the accuracy of pitch detection. Dynamic bit allocation is used according to different U/V modes of super-frames so as to reduce bit rate as well as improve speech quality. Subjective listening test results indicate that the speech quality of this algorithm is close to the existing federal standard 2.4 kb/s MELP coder.