Applications of Machine Learning in Speech Recognition

Abstract

As machine learning models have advanced, speech recognition systems have
become increasingly common. Virtual assistants, transcription software, and
automated customer support are now powered by these systems. Performance,
accuracy, and flexibility have increased with the use of machine learning
techniques including Recurrent Neural Networks (RNN), Deep Neural Networks
(DNN), and Hidden Markov Models (HMM). The main mathematical ideas
underlying these models are examined in this work along with an example
Java-based implementation and an analysis of current issues such data limits,
speaker variability, and noise reduction. Future options for enhancing voice
recognition with cutting-edge methods like transformer models and
unsupervised learning are discussed in the paper’s conclusion.