A Linear Prediction Coding On Perceptually Weighting

• Main Authors: Wei-Yen Lin, 林威延
• Other Authors: Jia-Yin Wang
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/98x355

碩士 === 中原大學 === 電機工程研究所 === 93 === Linear prediction coefficients (LPC) analysis is a commonly used technique in speech coding. It is used to describe the short-term correlations between speech samples. Another commonly used feature for speech coding is noise-masking. A speech signal with a higher intensity will mask another one with smaller intensity. Therefore, the speech with smaller intensity is more sensitive than speech with higher intensity to noise. In this thesis, we combine the conventional LPC analysis with the noise-masking feature. In the analysis process, the difference of the estimated signal and the original signal is weighted by an A-law or μ-law like function of the original signal in the minimization of mean squared error calculations. The proposed method is used in a code excitation linear prediction (CELP) coder for simulation test. After some simulations, we find the proposed method can achieve some improvement of the speech quality. Keyword: Linear prediction, LPC, weighted LPC