Analysis of prosody based automatic LID systems

Living beings inherently have the ability to differentiate languages as a part of human intelligence. Language Identification (LID) had been a science fiction in 1970's but today; it has been deployed in practical usage. The prosodic features of a speech are relatively simpler in their structure and are accredited to be very affective in some Language Recognition (LR) or LID tasks; irrespective of these features to be biased on numerous factors, as speaker's way of speaking, the culture and background of speaker. As because prosodic features is regardless very important, researchers against their heavy work have proven many methods for its normalization, making the feature inventory very large. In this book, we have used the maximum likelihood detector for the GMM-UBM based language model to analyze and identify various prosodic attributes in the LID tasks The entire wok includes a series of experiments on several speech corpus and different classification or/and identification technique. In an overview, we may assert that the book explores various experimental datasets, for, performance analysis of LID system with News speech and Natural Conversation speech and Joint Factor Analysis for LR on prosodic featured models.