IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
Differentiating human voice from other sounds like horn, tapping, bird
ksam917
post Nov 28 2013, 07:17
Post #1





Group: Members
Posts: 4
Joined: 16-November 13
Member No.: 112238



Hi all :-),

I am currently working on an academic project to find the human speech in the varying SNR conditions. I have implemented a method (using SNR band energy and SNR peaks in frequency domain) and is working fine to detect the voice activity but failing to detect only human speech. Currently I am failing to detect only human speech. I have tried with the speech feature extraction, but not able to make the decisions as thresholds values are varying for different environments. sad.gif

Please do some one suggest how to detect only human voice activity. Any suggestions will be very helpful.

Thank you,
ksam917
Go to the top of the page
+Quote Post
probedb
post Nov 28 2013, 09:25
Post #2





Group: Members
Posts: 1208
Joined: 6-September 04
Member No.: 16817



If it's an academic project you should have access to many relevant journals and papers on the same or similar subjects which will get you started.
Go to the top of the page
+Quote Post
Martel
post Nov 28 2013, 09:50
Post #3





Group: Members
Posts: 553
Joined: 31-May 04
From: Czech Rep.
Member No.: 14430



Just a hint - thresholds should not be absolute, they should be adaptive/relative. Absolute thresholds work only in a well defined/controlled environment. Have you tried some sort of normalization of the extracted features?


--------------------
IE4 Rockbox Clip+ AAC@192; HD 668B/HD 518 Xonar DX FB2k FLAC;
Go to the top of the page
+Quote Post
ksam917
post Nov 29 2013, 05:21
Post #4





Group: Members
Posts: 4
Joined: 16-November 13
Member No.: 112238



QUOTE (probedb @ Nov 28 2013, 10:25) *
If it's an academic project you should have access to many relevant journals and papers on the same or similar subjects which will get you started.


I have access to Journal, but not finding the suitable ones, and confused with from where to start and ow to proceed.
Go to the top of the page
+Quote Post
ksam917
post Nov 29 2013, 05:31
Post #5





Group: Members
Posts: 4
Joined: 16-November 13
Member No.: 112238



QUOTE (Martel @ Nov 28 2013, 10:50) *
Just a hint - thresholds should not be absolute, they should be adaptive/relative. Absolute thresholds work only in a well defined/controlled environment. Have you tried some sort of normalization of the extracted features?


Yeah I agree with you. The set thresholds are varying for the different environments. if we do normalization also these values will keep on varying right. How to set the adaptive thresholds?
Go to the top of the page
+Quote Post
Martel
post Nov 29 2013, 08:04
Post #6





Group: Members
Posts: 553
Joined: 31-May 04
From: Czech Rep.
Member No.: 14430



Consult your academic supervisor. Doing an academic project without any sort of supervision/leadership sucks and is a huge waste of time, in my opinion (been there, done that).


--------------------
IE4 Rockbox Clip+ AAC@192; HD 668B/HD 518 Xonar DX FB2k FLAC;
Go to the top of the page
+Quote Post

Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 21st August 2014 - 13:21