01-10-2010, 03:50 PM
[attachment=4760]
Software implementation of coherent demodulation of audio (speech) signals
Sharad Kumar
Student Id – 4114356
EE 4541 Digital Signal Processing Project
FINAL REPORT
abstract
INTRODUCTION
Demodulating a signal decomposes it into a slowly varying temporal envelope and a rapidly varying cosine (or fine structure--related to the instantaneous frequency of the signal). The motivation in representing a signal by envelope and fine structure comes from our desire to understand and model the signal processing function performed by the human ear, particularly the cochlea. The cochlea has been shown to decompose acoustic stimuli into different frequency components along the length of the basilar membrane (tonotopic decomposition). Also, the nerve fibres emanating from a high-frequency region in the cochlea convey information about the envelope modulations in the signa Thus, analytical signal models that characterize the envelope and instantaneous frequency information of a signal can be an effective tool in understanding how the cochlea functions. A commonly used method to analyze or modify the modulators of a broadband signal is to separate the signal into narrowband frequency subbands, and to decompose each subband into a carrier and a modulator. Several such methods are available in the literature, for example, the Hilbert envelope approach , which is also referred to as the “incoherent” approach and the “coherent” demodulation approach that is based on a carrier estimator For this project, analytic signal models based on the above two approaches, “incoherent” demodulation and “coherent” demodulation, were developed in Matlab. Two different kinds of filterbanks were used for the model, uniformly spaced (between 0 and Nyquist frequency) linear-phase FIR filter banks and one-third octave filterbanks, which more closely model the auditory filterbanks. The weakness of the “incoherent” model that it does not bandlimit the modulators and carriers was shown using speech signals taken from modulation toolbox (http://isdl.ee.washington.edu/projects/m...ontoolbox/). Then, it was shown that “coherent” demodulation did a better job at band limiting the envelopes. The effect of reducing the degree of temporal fluctuations present in a signal was also briefly studied.