论文部分内容阅读
With the development of telecommunication technology, the quality of multimedia data is getting more and more attention. Voice over IP (VoIP) is the destined trend of speech communication technology; however, it suffers from the delay and packet loss problems caused by the mechanism of Ethernet data transfer. Packet loss compensation (PLC) algorithms become increasingly important in VoIP, which effectively improve the speech quality in bad network conditions. Nevertheless, there is not any efficient method and standard to evaluate the performance of different PLC algorithms.The reconstructed speech quality is the direct manifest of the PLC algorithm. The commonly used speech quality evaluation methods are classified into two categories:objective and subjective test. The subjective test is directly related to the human ear perception, which leads it to the most effective evaluation method. However, it suffers from the big burden of repetitive listening test. The objective test is fulfilled by comparing the original speech signal and the reconstructed speech signal using some psychoacoustic model. It can be easily implemented, but its effectiveness, compared with that of the subject test, is a matter to be concerned.It this thesis, different possible evaluation methods for PLC algorithms are summarized, and an objective test method based on perceptual evaluation of speech quality (PESQ) is proposed. The PESQ value distribution is given based on the evaluation of large data collection. Furthermore, the comparison between the proposed objective test method and the subjective method is conducted, which proves the effectiveness of the proposed objective method. In addition, some description words are summarized from the subjective test to facilitate the improvement of PLC algorithms.