Detection of social speech signals using adaptation of segmental HMMs

This paper proposes an approach to detect social speech signals by computing segmental features using adaptation of segmental Hidden Markov Models (HMMs). This approach uses segmental HMMs and model adaptation techniques such as Maximum Likelihood Linear Regression (MLLR) and Maximum A Posterior (MAP) in order to acquire specific (or adapted) segmental HMMs that are fine-tuned to detect local regions of social signals such as laughter and fillers. Several segmental features are computed on automatically segmented audio with the specific segmental HMMs. Subsequently, the segmental features are used to detect social signals using Support Vector Machines (SVMs). The results indicate that the proposed segmental features play a significant role in detection of social speech signals.