Effect of speech segment samples selection in stutter block detection and remediation

摘要

Speech remediation by identifying those segments which take away from the substance of the speech content can be performed by identifying portions of speech which may be deleted without diminishing from the speech quality, but rather improving the speech. Speech remediation is important when the speech is disfluent as in the case of stuttered speech. We describe two stuttered speech remediation approaches based on the identification of those segments of speech which, when removed, would enhance speech understandability in terms of both, speech content and speech flow. We adopted two approaches, in the first approach we identify and extract speech segments that have weak semantic significance due to their low relative intensity; we subsequently trained several classifiers using a large set of inherent and derived features which provided a second layer filtering stage. The first approach was effective but required a two step process. In order to streamline the detection and remediation process, we introduced an enhancement which expands the realm of disfluency detection to include a broader range of speech anomalies by eliminating the need for a domain-dependent pre-qualification stage. The results of the new approach offer improved accuracy with enhanced simplicity, flexibility and extensibility.