[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
I need help in my masters thesis -- please read
- To: AUDITORY@xxxxxxxxxxxxxxx
- Subject: I need help in my masters thesis -- please read
- From: gehan kamel <gehan84@xxxxxxxxx>
- Date: Wed, 5 Apr 2006 04:53:00 -0700
- Delivery-date: Wed Apr 5 08:09:49 2006
- Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=Message-ID:Received:Date:From:Subject:To:MIME-Version:Content-Type:Content-Transfer-Encoding; b=5tbEFHiz/etkvvb71Fo2GvDphe+QN3M36Lq0mrmcNmy9YSfqWq76s46AI6cBf4xKWJiGsB9O83/FDKithfZ/pWyq574bdnYqvejhB/qrrZPRFFeGSaLng9tF0gtanNJ9TbenpWcY3o1E0lFrIKS70+HbSbtUQyMjKp28Vxt58kU= ;
- Reply-to: gehan kamel <gehan84@xxxxxxxxx>
- Sender: AUDITORY Research in Auditory Perception <AUDITORY@xxxxxxxxxxxxxxx>
My Name is Gehan Mustafa, I am a masters student and a teacher assistant in Cairo University (Egypt), Faculty of Computers and Information, IT Department. My Masters Thesis is a about Cocktail Party Speech Segregation( CASA and BSS models). If you don?t mind I have some questions concerning the topic:
1. The TDOA (Time Delay Of Arrival) of each time/ frequency region : how is it calculated ? What is TDOA, i.e. what does it represent when calculated for each time/ frequency region?
2. How is the local relative level or SNR calculated a priori for each time frequency regions?
3. If we use in CASA models the relationship between TDOA and the local relative level (RL, which is known a priori), this means we need to have the 2 sources unmixed to calculate the local relative level. Does this mean that such an algorithm is non-realistic to be tried out since in real world applications, no one has a priori the unmixed sources?
4. In my masters thesis, is it more realistic to work on the ST-Numbers Database or should I use a database which takes into account the echoing and reverberation effects, for the sake
of being more realistic and more applicable to real world situations which normally does contain echo and reverberation.
5. Do you advise me to work with CASA models or with BSS (Blind Source Separation) models? Which is more realistic for real-world applications? I think that BSS models are more realistic and more appropriate for me to use in my masters thesis. Do you agree with me or do you have another opinion?
I would really appreciate it if you could help me out in the formerly stated points. Thank you very much.
Yahoo! Messenger with Voice. Make PC-to-Phone Calls to the US (and 30+ countries) for 2¢/min or less.