| CLSP Homepage : Workshop Homepage | |
| |
| Workshop 2006 Calendar | Monday, November 23, 2009 |
|
Undergraduate Brown Bag Lunch 12:00 pm Barton 320A A meeting of all WS06 Undergraduates, WS and CLSP faculty from noon to 1 p.m. |
|
Dealing with Noise: Spectral Estimation, Beamforming, and Video 1:30 pm Shaffer 100 These lectures will organize classical and new ideas for acoustic noise amelioration. Specifically, the goal of these lectures is to cover the noise amelioration methods currently being tested for HMM and DBN recognizers operating on the AVICAR corpus. Lectures will be limited to one hour each; slides will be posted on the web afterward, with citations. Lecture 1: Spectral Estimation The classical tasks in auditory scene analysis are: 1 detect when something new happens, 2 figure out what happened. It turns out that, in order to answer question 2 as accurately as possible, it would be in the brainstems best interest to compute something like a Bayesian minimum-MSE estimate of the denoised perceptual loudness spectrum of the new event. Its pretty easy to write down the MMSE estimator of any function of a random signal in noise, but it can sometimes be difficult to actually evaluate it. The integrals can be solved analytically in just a few cases: specifically, if speech samples are assumed to be independent, then the MMSE estimators of the complex spectrum, spectral amplitude, and log spectral amplitude can be analytically integrated. Estimators of other functions e.g., PLP, or based on other speech sample distributions e.g., Laplacian, can be solved using numerical integration. Given any particular belief about the speech and noise PDFs, task #1 detect when something new happens can be written as a statistical hypothesis test. |
|
Voluntary Dinner Outing 7:00 pm Helmand Restaurant www.helmand.com |
Return to Calendar |