How to calculate Time stamp

Question

Mukesh Vaidya 2012-11-2

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/52564-how-to-calculate-time-stamp

Hi All, I have audio file i.e .wav. I have read this file using wavread...After that plotted a graph between time and amplitude. Now, I have to calculate the time stamp at which the amplitude changes first (i.e. where the vocal response starts fist). This audio file contains more than vocal responses. Difficulty is how to find best way to calculate the occurrence of first voice in each case. Can you please suggest me code.

2 个评论
显示无隐藏无

Jan 2012-11-2

编辑：Jan 2012-11-2

Please use a more meaningful title to the question. You want to locate the time of a voice, but this is not a "time stamp".

What have you tried so far?

Mukesh Vaidya 2012-11-5

Hi Jan,till now I have written a function which is calculating the reaction time when there is only one stimuli and its response but this program fails when there is more than one response in audio file thing is that it is also contains i.e. audio file fmri scanner noise. I am sharing you with my function, Please help me how to improve this. Thanks

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Image Analyst 2012-11-2

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/52564-how-to-calculate-time-stamp#answer_64087

编辑：Image Analyst 2012-11-2

在 MATLAB Online 中打开

From the help:

"[y, Fs] = wavread(filename) returns the sample rate (Fs) in Hertz used to encode the data in the file."

So can't you do something like this

firstLoudElement = find(y > thresholdValue, 1, 'first');
itsTime = firstLoudElement / Fs;

I'm not sure because I don't deal with sound signals very often. Of course you would have to define thresholdValue based on your signal and what you consider to be louder (a signal with higher amplitude) than the background sound. Of course if you have loud background sounds, like music, and want to find just where the voices start, ignoring the music, then you will have to get a signal where you have just voices first, like using frequency filtering, or something fancier and more effective like ICA:

http://research.ics.aalto.fi/ica/icademo/

http://research.ics.aalto.fi/ica/fastica/

http://cnl.salk.edu/~tony/ica.html