Deep Learning

CogniTech leverages the recent breakthroughs in deep learning in developing smart models trained at different levels of abstractions. For example, our team has developed deep-learning based keyword spotting and speech recognition engines as next generation interfaces for smart devices with varying levels of memory footprint. Our patent pending deep learning-based keyword spotting engine has small footprint and can detect accurately and in real-time the presence or the absence of target keywords in spoken utterances in clear as well as in noisy environments. It doesn't need to be trained by the user. The system developed by CogniTech is unique and makes use of many millions of elementary acoustic components to build the keyword. 

CogniTech is also pioneering development of anomaly detection based systems in streaming signals collected from devices and machines in industrial environment. This is done through the building of powerful predictor models based on deep learning tools. The technology has been proven successful in a number of case studies involving streams of data collected from real world industrial plants.