Feature analysis and neural network-based classification of speech under stress

It is well known that the variability in speech production due to task induced stress contributes signiicantly to loss in speech processing algorithm performance. If an algorithm could be formulated which detects the presence of stress in speech, then such knowledge could be used to monitor speaker state, improve the naturalness of speech coding algorithms, or increase the robustness of speech recognizers. The goal in this study is to explore which speech features are better stress relayers using a previously established stressed speech database (SUSAS). It is suggested that additional feature variations beyond neutral conditions reeect the perturbation of vocal tract articulator movement under stressed conditions. Given a robust set of features, a neural network based classiier is formulated based on an extended delta-bar-delta learning rule. Performance is considered for the following three test scenarios: mono-partition (non-targeted) and tri-partition (both non-targeted and targeted) input feature vectors.