Thread: Speaker Identification job interview

Hello,


Next week i'm going for a Speaker Identification job interview in the forensic department. I asked the one who called me how should i prepare for the interview and he told me to revise on topics in the DSP course (i have Bsc in E.E).
He didn't mention any specific topics.
I assume I should overview filters, sampling, maybe transformations, but I don't know exactly what.
I'm really interested in this job and after i googled about it i came with your site.
could you please shed a little bit more light on the topics i'll have to read and practice ?
which questions may the interviewer will ask me(professionally) ?


thank's in advance,
Good day
Ehud

You might want to read up on formants (if you haven't already).
Formant - Wikipedia, the free encyclopedia

That's about the limit of my knowledge of the subject. Good luck!

Fourier series and FFT (Fast Fourier Transform) could be usefull tools.

Hi Ehud, my name's Chris, I've spent several years running an employment company helping to teach people how to organise others to create jobs in areas where existing jobs are hard to come by. During this time we also taught people how to teach others to prepare for job interviews, whilst I wasn't personally too heavily involved in this side of things I did gain some experience along the way in conducting interviews.

Whilst I don't profess to have any detailed knowledge about the subject area in which you are applying I can possibly point out some of the basic areas of knowledge you may find your interviewer wishing to examine you on.

Such as: "Please explain your views on the GMM-UBM APPROACH?"

Or perhaps: "What would you consider a key advantage of using THE MIXED GMM AND SVM APPROACH?"

They may ask: "How do you feel about the training, the length of time it may last and benefits?"

Possibly: "How would you describe the caution that should be exercised when applying speaker recognition techniques?"

It should be noted as I stated above that I don't have any specific insight in this area and these are some just some basic questions that I might have chosen to test your knowledge.

I would suggest however that prior to your interview you go and speak with people working in this field to get some first hand knowledge of what it may entail, this will be very helpful for your preparation and give you confidence. I would further recommend you try to read up on a least a little bit of each aspect this way you shouldn't be caught out and then also comprehensively prepare a detailed answer on a key subject, in this way regardless of what you are asked you can lead the conversation on to your prepared answer.

I would also suggest this is the way you approach your interview, as an opportunity to discuss a subject you have researched and are really passionate about. If you can, try and imagine that your interviewer is already your friend, but don't be to informal, this will help create a natural rapport. Remember the more your interviewer likes you the more likely you are to get the position, also all interviewers like people who are passionate about their subjects.

Finally remember that an average person's best is usually better than a really good person's average.

Best of luck and I hope you get the position.


P.S.

Resources:

W. Campbell, D. Sturim, D. Reynolds, and A. Solomonoff, “SVM-based
speaker verification using a GMM supervector kernel and NAP variability compensation,”
in Proc. Int. Conf. Acoustics, Speech, and Signal Processing, 2006

J.-F. Bonastre, D. Matrouf, and C. Fredouille, “Artificial impostor voice
transformation effects on false acceptance rates,” in Proc. Interspeech, 2007

G. R. Doddington, W. Liggett, A. Martin, M. Przybocki, D. Reynolds, “Sheep,
goats, lambs and wolves: A statistical analysis of speaker performance in the NIST
1998 speaker recognition evaluation,” in Proc. Int. Conf. Spoken Language Processing

S. E. Mezaache, J.-F. Bonastre, and D. Matrouf, “Analysis of impostor tests
with high scores in NIST-SRE context,” in Proc. Interspeech, 2008.

N. Brümmer and J. du Preez, “Application-independent evaluation of speaker
detection,” Comput. Speech Lang, 2006.

R. Schwartz, “Voiceprints in the United States—Why they won’t go away,” in
Proc. Int. Association for Forensic Phonetics and Acoustics, Göteborg, Sweden.

I would follow the advice to review your DSP materials, so focus mainly on that — particularly spectrum analysis, 2D- and 3D-representations of speech, etc. Interviewing you as a BSEE, the interviewer is probably looking for something along the lines of Fourier series and FFT as mentioned above. It sounds like you might also be matching unknown spectrums to known spectrums — that is, matching a sample of unknown speech/spectrums to a library/database of known speech/spectrums, giving the highest probability to the lowest. This is done in other forensic work, such as GC/MS results in chemistry.

It would probably be somewhat helpful to know a little about the frequency spectrum of sound that generally represents a vocal sound, but I wouldn't try to memorize anything like a "v" is a called "voiced labio-dental fricative" (because there's probably 100 or more different vocal sounds, and it's just not worth the effort). It's more important to know the general spectrum that a "v" produces, but again, the interviewer didn't emphasize or even mention this.

If you're interested in knowing a little about speech production, take a look at this, and you'll see general speech types, such as plosives, nasals, fricatives, etc and that some sounds are very similar (P versus B, T versus D, K versus G, F versus V, etc), where one is "unvoiced" (vocal cords don't vibrate) and one is "voiced" (vocal cords do vibrate). So, if your interviewer mentions a "plosive", you'll know what it means.

And searching for general info on the web, "speaker identification" might also be called voiceprint identification, if I understand you correctly.

Hi ,

thanks strange,cakus & Chris

as for the basic questions Chris the first two aren't too basic as I mentioned I was told to prepare on specific area and for the more general questions I also did a little research and I'm all the time asking myself what he would ask me like:
why would you prefer a low income job in the police instead of working in an Hi-tec company ? also about the training , maybe on being recruited to be in the police force...

strange & cakus - I doing again my DSP homework assignments...

anyone else maybe on analyzing block diagram, noise reduction...