| The technique of extremely-low speech coding is the basic of modern secret communication. The requirement of extremely-low speech coder is on the increase as the rapid development of modern secret communication. In order to meet the requirement of secret short-wave narrow bandwidth data communication, and to improve the anti-disturbance and real-time features of the battle-field communication, it is necessary to use the extremely-low speech coding at a rate below 2400bit/s. This paper makes an intensive study of the chief method and technique of extremely-low speech coding.This dissertation study on the Lpc-10 vocoder speech encoding techniques, and the normalized cross-correlation function pitch detection algorithm(NCCFPDA), which greatly boost the accuracy of the detection of voiced and unvoiced speech, is introduced to improve the judgment error own to pitch multiplication-frequency and half-frequency. Using the quantified feature of the linear spectrum pair(LSP) parameters, the algorithm reduces the error rate of the signal transmission and improves the accuracy and efficiency of the quantification. By introducing a fixed vector quantization codebook (LSPVQ), the algorithm improve the accuracy and efficiency of the quantification and reduce the transmission rate. By introducing the super-frame of 4×22.5ms, the algorithm makes a better balance between the confliction of low-bit rate and serious deterioration of voice quality. Combined with parameters interpolation, the algorithm greatly reduces the transmission rate. By using these coding technologies, the algorithm at the 900 bit/ s rate has a relatively satisfied synthetic voice, and is realized in MATLAB. The simulation results show that the synthetic voice is understandable and clarity as the traditional 2400 b/s Linear Prediction LPC-10 speech coding algorithm.Compared with the existing algorithm, this algorithm is simple for operation, good quality, practical, and a promising low rate speech coding method. |
PDF Full Text Request