Voice MFCC feature extraction (voice recognition)
2016-08-23
31 0 0
4.2
Other
Earn points
In the processing of speech signals, basically using short-acoustic parameters. Acoustic parameters of so-called short-time was consistent with 20~40 Ms a frame processed voice, by means of Fourier Transforms, and then after dimension discrete cosine transform of a feature. This feature is generally chosen is the hundreds of voice sampling points (20ms*8K=160), the output is a fixed dimension 39~57 dimension parameters, used for pattern recognition. MFCC is an auditory frequency cepstrum parameter, the parameters from the frequency of sound to the human ear's nonlinear psychological feelings reflect the speech characteristics of short-time magnitude spectrum, so in both speech recognition and speaker identification has been in an extremely wide range of applications.
This procedure for extracting the MFCC feature parameters of the voice files, input speech SPH format files, output-suffix parameters called MFC files. Taaa.SPH is the original audio file
This procedure for extracting the MFCC feature parameters of the voice files, input speech SPH format files, output-suffix parameters called MFC files. Taaa.SPH is the original audio file
c++
语音
识别
mfcc
提取
参数
特征
Related Source Codes
Chinese Speech Recognition
0
0
no vote
Select features with high correlation
0
0
no vote
Local Path Planning Algorithm - DWA Algorithm
0
0
no vote
enDAQ-Shock-Data-Share-SRS-Blog
0
0
no vote
Calling chatGPT in a Windows application
0
0
no vote
No comment