1 / 3

Feature extraction

Feature extraction. 在理論上求取 MFCC 的流程. 語音訊號. 取 30ms 為一個音框. frame. frame. H(n)-H0.95H(n). Pre-emphasis. 抑制 sidelobe 部分的信號,把 peak 凸顯出來. Hamming. 轉成頻率 domain 較易分析. FFT. 求振幅. || ||. Filter Bank (Triangle). Mel (f)=2595 log (1+f/700). 10. log. Inverse Cosine Transform. 求梅爾參數.

Télécharger la présentation

Feature extraction

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Feature extraction

  2. 在理論上求取MFCC的流程 語音訊號 取30ms為一個音框 frame frame H(n)-H0.95H(n) Pre-emphasis 抑制sidelobe部分的信號,把peak凸顯出來 Hamming 轉成頻率domain較易分析 FFT 求振幅 || || Filter Bank (Triangle) Mel (f)=2595 log (1+f/700) 10 log Inverse Cosine Transform 求梅爾參數 features

  3. begin • Initial MFCC • hamming table • FFT table • triangular filter bank coefficients filter • find Low, upper freq and space • find center freq • set response for every filter • all channels • set upper and lowerside spk eof 1 0 • Open file • fea • len • dc bias utt utt>Nutter 1 0 • open vat file • utterance count ++ • Read head of vat file • read 256 byte header • get sample number If status = 1 ? (check condition and transcription of waveform) no yes • compute total frame Extract MFCC • save feature out to • file • close fea , len • file end

More Related