Home Sound: A GPU-based Platform for Massive Data Acquisition and Processing for Acoustic Ambient Assisted Living Applications for Behavior Monitoring

Abstract

Human life expectancy has grown over the last century, which has driven governments to increase the efforts on caring about the eldest population. Therefore, modern trends take advantage of latest advances in technology to remotely monitor those people with special needs at their home, increasing their life quality and with less impact on their social lives. This paper presents an acoustic event detection platform for assisted living that tracks patients’ status by automatically identifying and analyzing the acoustic events happening in a house. Specifically, we have taken benefit of a Jetson TK1, with its NVIDIA Graphical Processing Unit, to process the acoustic data and identify a closed number of events in order to inform the care system. This is a proof of concept conducted with data of only one acoustic sensor, but we plan in the future to deploy a sensor network in several places in the house.

Keywords

• Ambient assisted living
• Sensor network
• Machine hearing
• Acoustic feature extraction
• Machine learning