Sprecheridentifizierung mittels SVM bei der Oriya-Spracherkennung

Many contributions have been made by language processing scientists and researchers around the world for both English and non-English languages. However, there has been very little research conducted on the spoken Oriya language so far. This research work is a completely new contribution to the Oriya language, which is one of the recognized official languages of India. Several aspects are considered in this work. Statistical pattern recognition algorithms are employed in the development of speaker identification and Oriya speech recognition. The speaker identification is based on the Support Vector Machine (SVM) algorithm, while the Hidden Markov Model (HMM) algorithm is used in the development of the Oriya speech recognition engine. Other effective techniques such as Dynamic Time Warping (DTW), ontology-based Backpropagation Feed Forward Neural Network (BPFF), k-neighborhood, etc., are also examined and analyzed in terms of speaker identification and Oriya speech recognition performance. A speech corpus is the integral unit of any language processing application. Since no publicly available Oriya speech corpus exists, we have created two different speech corpora in Oriya.