Deep Learning Based Lip Reading for Speech Recognition

Automatic speech recognition (ASR) is the process of using machine learning techniques to convert human speech into written text. This paper proposes a deep learning-based lip reading system, which can be beneficial for individuals who are unable to speak and can also identify what a speaker is saying in noisy environments. Lip reading is a visual technique that recognizes speech based on the movements of the lips. The paper discusses various components of the proposed lip reading system, including lip detection, feature extraction, feature modeling, and sentence prediction. Additionally, a comparative analysis was conducted to evaluate the accuracy of different deep learning models in recognizing the spoken sentences. Conventional speech recognition systems have been found to be less efficient compared to deep learning-based systems. Over time, traditional lip reading systems have been replaced with their deep learning counterparts, as the deep learning approach can leverage a large amount of data to achieve much greater accuracy.