Classification of speech recognition by using sequential minimal optimization algorithm

The categorization and recognition is a very recent development in the realm of machine learning. This study shows the categorization of emotions using the architectural framework of a Distributed Speech Recognition System (DSRs), accompanied with the related results of performance evaluation. The temporal patterns of semantic units, such as sentences and words, characterized by using a set of 3800 statistical factors. The use of the KDDM (Knowledge Discovery and Data Mining) program was employed to conduct the procedure of determining the most pertinent components for classifying emotional states. Subsequently, a thorough analysis was performed on the data obtained from various classification methodologies. The findings, derived from the analysis of the California Database of Emotional Speech and the Actual Stress corpus and Speech Under Simulated, indicate that the optimal outcomes are attained by employing a Sequential Minimal Optimization (SMO) algorithm to feeding and training the Support Vector Machine (SVM). The aforementioned result is achieved by the normalization and discretization of the statistical parameters given as input.