A Super-Resolution Spectrogram Using Coupled PLCA

Date: 
Fri, 10/08/2010 - 3:15pm - 4:05pm
Event Type: 
DSP Seminar
This week's DSP seminar will be presented by Juhan Nam, a CCRMA Ph.D. candidate researching audio signal representations and digital emulation of analog waveform generators.  The seminar will take place in the CCRMA classroom (Knoll 217) at 3:15 PM this Friday, October 8.  Abstract follows:


A Super-Resolution Spectrogram Using Coupled PLCA

The short-time Fourier transform (STFT)-based spectrogram is commonly used to analyze the time-frequency content of a signal.  By adjusting the window length used, the STFT provides a trade-off between time resolution (which is inversely proportional to window length) and frequency resolution (which is proportional to window length). This work presents a novel method that achieves high resolution simultaneously in both time and frequency.  The idea is to construct an underlying high resolution spectrogram, which may be smoothed along the time axis to produce the standard long window length (high frequency resolution) spectrogram, or smoothed along the frequency axis to produce the standard short window length (high time resolution) spectrogram.  This is achieved by extending Probabilistic Latent Component Analysis (PLCA) to jointly decompose the short window length and long window length spectrograms to generate a new spectrogram, maintaining high resolution in both time and frequency. Termed the “super-resolution spectrogram”, it can be particularly useful for speech as it can simultaneously resolve both glottal pulses and individual harmonics.

Open to the Public
Syndicate content