| Summary: | 碩士 === 國立暨南國際大學 === 資訊工程學系 === 89 === The main topic in this thesis is to investigate the effectness of speech communication quality when any packet loss occurs in a network system. Furthermore, a feasible method to reconstruct the lost speech information containing in the lost packets is presented. The codec, which is designed follows the standard FS1016 4.8kbps is implemented in order to study the quality of speech communication over a network. A code excited linear prediction (CELP) codec is implemented according to the specification of the standard. Based on this implemented codec, we not only study the quality of speech communication on networks of various packet loss rates, but also present a new method to reconstruct the speech information contained in the lost packets. The speech information is reconstructed via recalculating the parameters of speech frame originally stored in the lost packets. In order to have a better quality after reconstruction, one may extract some information from a given speech frame and preserve this information in preceding speech frame and successive speech frame of the given frame; however, this method will unavoidably increase the bit rate of codec. In order not to increase the bit rate of the codec, a new method of reconstruction of the speech information is proposed in this thesis. By computer simulations, this new method results in an acceptable speech quality, but does not increase the bit rate of the codec.
|
|---|
