Automatic Speech Recognition
MIT's graduate course on building systems that convert speech into text. The curriculum moves through three parts: acoustic theory of speech production, acoustic-phonetics, and signal representation; then algorithmic foundations including pattern classification, search algorithms, stochastic modelling, and language modelling; then advanced topics comparing recognition approaches, robust recognition under noise, speaker adaptation, paralinguistic processing, and multimodal systems. Materials come from MIT OpenCourseWare and include lecture notes and assignments covering the mathematics and engineering behind modern speech recognition pipelines. The course assumes background in signal processing and probability, and is aimed at students who want to understand how systems like dictation software and voice assistants are actually built, not just how to call an API. No certificate is offered through OCW; materials are free to use under a Creative Commons license.