Kolodij R. Development of methods and facility for efficiency improvement of speech signals compression in telecommunication systems for IP-telephony.

Thesis is devoted for development of methods and facility for quality improvement of compressed speech signal for application into IP-telephony networks based on the methods of speech signal wavelet-conversion and its coding with changeable bit-rate. The structure of coder using the method of wavelet-conversion, which includes all the necessary component for work into IP-network: changeable coding bit-rate, adaptation on VAD, selecting of various types of wavelets and noise-stable coding and minimum values of calculating and temporal expenses, was developed. For evaluation of coding quality the simple enough methodology of relationship determination between MOS i MSE was proposed. Its results can be applied for determination of communication quality due to considering possibility of the most various network factors, especially, coding type, influence of buffer, delay, its irregularity and losses of packets. The developed method restores well a speech signal even at 30% losses of packets into network atany distribution law that was verified as a result of the simulation, as well as its results are correlated well with the knownone. The simulation shows that the developed method for coding of speech signal on the base of wavelet-conversion provides the coding of signal form. The achieved coefficient of speech file compression is 20, at that at VAD on level the 1/32 bit-rate does not exceed 4.1 kbps, the minimum value of MOS is equal to 3.83.