Continuous Automatic Speech Recognition System Using MapReduce Framework

摘要

Now-a-days, Speech Recognition had become a prominent and challenging research domain because of its vast usage. The factors affecting Speech Recognition are Vocalization, Pitch, Tone, Noise, Pronunciation, Frequency, finding where the phoneme starts and stops, Loudness, Speed, Accent and so on. Research is going on to enhance the efficacy of Speech Recognition. Speech Recognition requires efficient models, algorithms and programming frameworks to analyze large amount of real-time data. These algorithms and programming paradigms have to learn knowledge on their own to fit in to the model for massively evolving data in real-time. The developments in parallel computing platforms opens four major possibilities for Speech Recognition systems: improving recognition accuracy, increasing recognition throughput, reducing recognition latency and reducing the recognition training period.