The Research On Audio Quality Assessment

With the development of technology,more and more multimedia data come into our life.People usually prefer to enjoy the multimedia data with high quality. However, big data made it even more difficult for people to choose because of quality in different levels. Therefore, it is inportant to work on multimedia quality assessment.As to audio quality assessment, almost all the popular music sharing platform online only take the bitrate as the measure of audio quality. However, it is not reasomble because audio quality is very subjective and can be influenced by many fiictors. Only one or two sinple characteristics could not explain it Besides,felse quality audio makes it even more difficult to assess audio quality precisely. Neither subjective method nor objective method today can solve this problem properly. In this paper, different fee tors of effecting audio quality have been stidied. And according to different applications, a non-reference audio quality assessment method based on manifold ranking with SVR and another method aiming at felse quality assessment based on FW are proposed. Here are the research findings:(1)This paper reviews the multimedia quality evaluation, including the research background and significance, image quality assessment, video quality assessment, audio quality assessment. Then this paper focus on audio quality assessment and analyze human auditory system and psychoacoustics. Then subjective evaluation standard ITU- RBS.1116-1and objective assessment standard ITU-R BS.1387 are studied.(2)A more effective non-reference audio quality assessment method is proposed. The first method for non-reference audb qmlity assessment NFAQ was proposed in 2013.However,it only extracts features from audio and ignores the relationship between different audios. This paper take the relationsh:p between different audios into consideration and proposes a novel method based on manifold ranking and SVR. Then coirpare it with NFAQ.The result shows that the proposed method is more efiective,(3)A method is proposed to assess audio with felse quality. First the encoding and decoding of MPS is studied. This paper analyzes how the audio with felse quality is made.Then the audio with false quality can be revealed by estimating the FVV of quantized MDCT coefficients. Experiment shows that the performance of the proposed method is very good.