split wav file based on txt file (two vectors - start time and duration)

Hello,

I have a wav file (dialogue between mother and her son) , and i have already split the data for each speaker , meaning i have a txt file witch containing data (in seconds) about the start time and duration of speaking for each speaker ;

######

for example ,

mother.txt :

start time  duration
1.011	4.457
5.974	1.166
8.26 	8.616
17.493	5.009
24.688	7.303
35.011	2.388
38.71	6.061
45.993	3.898
51.73	12.096

#######

and i want to cut each part of the txt file from the wav file (creating wav file for each part) and merge them into one wav file( witch have all the words/sentences that the mother said in the dialogue )

which functions should i use ?

thanks a lot!

Answer 1

[Mathieu NOE]

hello Fatima

happy new year !

this code will do the trick :

[song,FS]=audioread('demo.wav');  
channel = 1;
song = song(:,channel);

% create list of start / stop time indexes 

% mother.txt  :
% start time  
start_time = [1.011,	
5.974,	
8.26 ,	
17.493,	
24.688,	
35.011,	
38.71,
45.993,	
51.73];

% duration
duration = [4.457,
1.166,
8.616,
5.009,
7.303,
2.388,
6.061,
3.898,
12.096];

stop_time = start_time+duration;

%  main loop % 
song_extract_complete = [];

for ci = 1:length(start_time)
    % then determine start and stop indexes , knowing the start and stop times
	song_start_ind = round(start_time(ci)*FS); % must be an integer value 
	song_stop_ind = round(stop_time(ci)*FS); % must be an integer value 
    % then extract
	song_extract = song(song_start_ind:song_stop_ind);
	song_extract_complete = [song_extract_complete; song_extract];
end

% export in wav format
audiowrite('demo_out.wav',song_extract_complete,FS);

[Fatima Abbas]

Happy New Year!

Thank you!!! ,this is exactly what I was looking for :)

[Mathieu NOE]

you're welcome !!

Answer 2

[dpb]

That only would take basic math operations to turn the start times and durations into indices into the the .wav file based on the sample rate of the file.

The index would simply be

iStart=tStart/Fs;

and the number of samples per duration

nSamples=tDuaration/Fs;

so each end time would be

iEnd=iStart+nSamples;

You may want to adjust the start time by +1 for time being zero-based while the MATLAB array indices begin at one, but one presumes the sampling frequency is something like 44 kHz so that adjustment is probably immaterial in practice.

[Fatima Abbas]

thanks for you answer,

how can I create a new wav file depending on the indices of the original wav file?

[dpb]

Just iterate over the computed indexing arrays and write those subsections of the input array to a new file; one for each subject.

Of course, having the indices means you could also just process the original file directly without the need to actually create a separate file, although there might be other reasons to do so.

Give it a try and see how far get...it's more educational to learn by doing than just have somebody hand over a solution that don't see how was done.