Analyzing sounds of home environment for device recognition

Abstract

Home environments are one of the subjects of study regarding ambient intelligent systems for various purposes, including development of assistance systems for the elderly and energy consumption optimization. Sensing the environmental state via different sensors is the first and crucial component of every ambient intelligent system. In this work we investigate the use of environmental sounds for touch-free audio-based device recognition in a home environment. For this purpose, we analyzed sound characteristics of typical home appliances using different processing techniques. We are using the acquired knowledge to develop a flexible set of features, which can be set manually or determined automatically. To classify the device-specific acoustic fingerprints - consisting of a significant subset of our features - we use established supervised learning techniques, whereby we optimized the straightforward ones. After building a recognition basis for the recognition of fixed length sound buffers on demand, we implemented a live recognition mode for real-time environment monitoring, providing runtime setup adjustments. We then extended our work with the recognition of untrained, simultaneously working, known devices by mixing their records, utilizing semi-supervised learning. We then anticipated promising results in our evaluation in various aspects, including recognition rate, performance for the different combinations of features, as well as to study the reliability of an automatic mixing of trained data.