Research And Implementation Of Narrowband Speech Quality Measurement Algorithm

From the main assessment, speech quality assessment can be divided into subjective and objective assessment. Subjective assessment is the real reflection on speech quality, but this method is time-consuming, expensive and can not be effectively repeated. To develop a speech objective evaluation system, which is flexible, convenient and reliable, is essential.PESQ(PerceptuaI Evaluation of Speech Quality) algorithm is put forward by Rix, Beerends and etc in 2000, which becomes ITU-T Recommendation P.862. Compared to other algorithms, it takes the end-to-end delay into consideration, which can evaluate kinds of networks. It also adopts improved acoustical and advanced technology, which is robust to the communication delay and noise of environment and is totally suitable for the communication system. The paper proposes a PESQ system based on ADI Blackfin533. The algorithm and codes are optimized to achieve assessment for narrowband speech quality in real time.The main work of this thesis includes the following parts: The first chapter introduces the different methods and performance of speech quality assessment. The second chapter introduces the earlier models and the weakness of objective assessment, at the same time, the popular applied speech objective assessment model-PESQ is analyzed in detail, then it makes an introduction to the DSP chip-Blackfin 533 and the development environment-Visual DSP++4.5. The third chapter proposes the scheme of software and hardware design as well as system structure. The last chapter focuses on the implement of PESQ arithmetic based on the hardware platform of DSP-BF533EZ-KIT Lite. The algorithm and codes are optimized and optimization methods are given in detail.