Speech Recognition Systems: A Comparative Review

摘要

Creating voice control for robots is very important and difficult task. Therefore, we consider different systems of speech recognition. We divided them into two main classes: (1) open-source and (2) close-source code. As close-source software the following were selected: Dragon Mobile SDK, Google Speech Recognition API, Siri, Yandex SpeechKit and Microsoft Speech API. While the following were selected as open-source software: CMU Sphinx, Kaldi, Julius, HTK, iAtros, RWTH ASR and Simon. The comparison mainly based on accuracy, API, performance, speed in real-time, response time and compatibility. the variety of comparison axes allow us to make detailed description of the differences and similarities, which in turn enabled us to adopt a careful decision to choose the appropriate system depending on our need.