The Research Of Speech Codec Based On DSP In Digital Intercom

With the development of economy, science and technology, people have greaterdemand for information resources. Thus, the amount of information that needsprocessing and transmitting becomes larger, which in turn calls for a higher level ofcommunication, such as, its safety, multi-service, convenience and the low powerconsumption of its wireless terminal. Hence, to transmit as much as informationthrough the fewest bands gets increasingly important in wireless communication. Forthe purpose of transmitting information efficiently, in a high-quality and secure way,it is necessary to transform source information signal at the transmitting end ormodulate signals on the way of transmission. Since the wireless band grows morevaluable and the communication service more diverse in the field of intercom, thetraditional analog intercom cannot meet the demand, it is unavoidable for the digitalintercom to replace the analog one. Recently, China has been developing the digitalintercom of its own property rights and the related protocol.Speech signal processing is critical technology of digital intercom. Whethervoice communication can be achieved on the regulated bandwidth depends on thereal-time modulation and demodulation of speech signals. Mixed Excitation LinearPrediction algorithm was chosen to deal with speech signals based on the researchand analysis of speech signal processing. The specific implementation includes threeparts. First, the implementation structure of MELP algorithm was simulated andverified in Matlab. Then, the hardware test of MELP algorithm structure was realizedon DSP hardware platform of TI. At last, the experimental data produced by DSP andMatlab was compared and analyzed. As the results showed in this thesis, MELPalgorithm could realize speech compression codec effectively.