Dynamic sensor activation and decision-level fusion in Wireless Acoustic Sensor Networks for classification of domestic activities

For the past decades there has been a rising interest for wireless sensor networks to obtain information about an environment. One interesting modality is that of audio, as it is highly informative for numerous applications including automatically classifying domestic activities that is focussed on in this work. However, as they operate at prohibitively high energy consumption, commercialisation of battery-powered wireless acoustic sensor networks has been limited. To increase the network's lifetime, this paper explores decision-level fusion, adopting a topology where processing -- including feature extraction and classification -- is performed on a (dynamic) set of sensor nodes that compute classification outputs which are fused centrally. The main contribution of this paper is the comparison of decision-level fusion with different dynamic sensor activation strategies that leverage the redundancy of information in the network. Our results show that representing the classification output using vector quantisation can reduce communication per classification output to 8 bit without loss of significant performance. In case of fixed sensor activation this results in an energy reduction up to 3%. While the savings of fixed sensor activation are limited, it is shown that dynamic sensor activation, using a centralised approach, can provide an energy reduction up to 80%. In general, this work indicates that if opted for a topology using decision-level fusion, dynamic sensor activation is needed when a long battery lifetime is desired.

Publication year:2021

Accessibility:Open