Research Description:

My research in Digital Signal Processing (DSP) at UF's Computational Neuro-Engineering Lab focuses on Time-Scale Modification (TSM) algorithms and their application to speech signals. The intent of Time-Scale Modification is the compression or expansion of an audio segment while preserving the original pitch. In particular, my research focuses on expanding a signal's time base using Fourier Transforms in the TSM algorithms. The intent is to test the hypothesis that prolonging certain parts of speech will lead to clearer cell phone conversations, particularly in noisy environments. Fourier Transforms play a pivotal role in DSP by expressing a signal in the frequency domain rather than the time domain and allowing extraction of a signal's composite frequencies. This required a mathematical analysis of Fourier Transform Theory reviewing the development of Fourier Transforms from the mathematically ideal Fourier Series to the practical Fast Fourier Transform (FFT). Because speech is a non-stationary signal, the FFT must be further modified resulting in the Short-Time Fourier Transform (STFT). My approach uses the STFT's bin phases to determine a signal's true frequencies in order to preserve them during processing and linear interpolation of the STFT to actually achieve the time scaling. This is currently being implemented in MATLAB. In addition, I have explored the possibility of interfacing C and MATLAB. Future goals include programming a DSP chip.