Analysis of Acoustic Feature Extraction Algorithms in Noisy Environments
Weiyang Cai · 2013
I would like to express my greatest gratitude to the people who have helped and supported me. First, I would like to thank Professor Wendi Heinzelman for her continuous guidance and invaluable advice on my thesis. Many thanks to my parents for their undivided support and encouragement. I would also like to thank Na Yang and He Ba from the Wireless Communication and Networking Group for providing advice on my thesis. This research was part of the Bridge project, which was supported by funding from National Institute of Health NICHD (Grant R01 HD060789).iv Acoustic feature extraction algorithms play a central role in many speech and music processing applications. However, noise usually prevents acoustic feature extraction algorithms from obtaining the correct information from speech and music signals. Thus, the robustness of acoustic feature extraction algorithms is an area worth studying. In this thesis, we consider two important acoustic features: pitch and speaking rate. For each acoustic feature, we introduce several classic and state-of-the-art feature extraction algorithms and evaluate the performance of each of them in noisy environments. We analyze the results and provide possible explanations why some feature extraction algorithms outperform the others in noisy environments. v