Improved Mixed Excitation Linear Prediction Coding Algorithm

How to transfer voice information effectively, is still the focus of study in many communication system with the development of science and technology.The earliest voice communication through the analog voice has realized the long-distance voice transmission, but the digital communication’s coming with the advantages in the aspect of anti-jamming ability, fast switching capability,reliability than the analog communication replaced the analog voice.but the digital communication’s drawback is occupies too much bandwidth. speech coding technology’s deep research reduces effectively the frequency band is occupied by the digital voice signal.In recent years, the problem of limited bandwidth seems to have been resolved, but with the development and the actual needs of digital communication, reduce the speech coding rate, improve the utilization ratio of the channel has become the urgent need of the people. So the high quality and low bit rate speech coding algorithm become essential research direction in communication.At present, Low bit rate speech coding has developed more complex in the two aspects of algorithm and structure, MELP algorithm is one of the best algorithm, is even considered to be the most potential algotithm.The algorithm is based on the traditional LPC model, replace the simple duality excitation with mixed excitation, using multi band ideas, better simulate the natural characteristics and high quality.This paper introduced two large improvement, in view of the complexity of pitch tracking and roughness of sub-band voice intensity,through analyzing the MELP algorithm codiong and encoding privciple. First the pitch extraction method is improved into a more simplified algorithm;second, Design a subband vector quantization method of sub-band voice intensity and achieve it.then get an improved MELP vocoder algorithm (A-MELP). In the end a large number of and full simulation experiment are done for the improved MELP algorithm(A-MELP).