M. Sigmund and T. Dostal (Czech Republic)
speech recognition, emotional stress in speech
Normal speech may be regarded as speech made in a quiet room with no task obligations. Stress in speech, on the other hand, is a result of speech produced under emotional states, fatigue, heavy workload, environmental noise or sleep loss. In this paper we investigate acoustic indicators for stress in speech. Various simple features such as vowel duration, formants, fundamental frequency, cepstral coefficients and long-time spectrum are applied in this study. New Czech database of stressed speech was created for use in our experiments consisting of data collected during oral final examinations at our university. Results of tests show that examination stress can be detected using selected speech features.
Important Links:
Go Back
