Development of an audio detection system using wireless sensor networks

Abstract

Wireless sensor networks offer a number of attractive advantages for the realization of many applications that were previously difficult to realize through the use of traditional sensors. This report presents details on a project which was aimed at deploying a wireless sensor network for the purpose of acoustic tone detection. The tone detection system was developed using a wireless sensor network based on Mica2 sensor nodes and a PC based processing system that was implemented on a Java platform. A number of hardware constraints were encountered and various optimization tests were conducted to determine the best possible parameters to achieve optimal performance. Sampling rate was a key constraint that limited the ability of the system and a number of techniques were used to maximize the sampling rate achievable by the sensor nodes. Through tests conducted to evaluate system performance, the ability of the system to detect single tones, multi tones, harmonic components and noise contaminated single and multi tones was demonstrated. The system was also capable of successfully detecting the relative power levels of these frequency components. A sample application for the tone detection system was also explored. A simple sound detection system was developed through this exploration, by enhancing the system with the required additional components. Though not rigorously pursued in terms of application specific details, the sound detection system that was developed was capable of successfully detecting the presence of a selected mechanical sound. All objectives were successfully achieved by the project. However, several enhancements and applications could be identified. Several areas for future study/work on these enhancements and applications were also identified.