Title: Robust Phonetic Feature Extraction Under a Wide Range of Noise Backgrounds and Signal-to-Noise Ratios
Abstract:A method for automatic classification of articulatory-acoustic features (AFs) and phonetic segments has been developed that is relatively immune to performance degradation under a wide range of acoust...A method for automatic classification of articulatory-acoustic features (AFs) and phonetic segments has been developed that is relatively immune to performance degradation under a wide range of acoustic-interference conditions. A key property of the classification method is to train on two separate noise backgrounds (“pink” and “white” noise) across a 30-dB dynamic range of signal-to-noise ratios (SNRs). This training regime reduces the error rate at the articulatory-feature and phoneticsegment levels by as much as 40-60% for low-SNR conditions relative to the baseline system (trained solely on “clean” speech) and thus ensures that phonetic-segment classification is sufficiently high (60-80% accuracy) as to provide reasonably robust word recognition performance at low SNRs.Read More
Publication Year: 2001
Publication Date: 2001-01-01
Language: en
Type: article
Access and Citation
AI Researcher Chatbot
Get quick answers to your questions about the article from our AI researcher chatbot