Embedded Sound Localization on an Axis Camera

Niklas Svensson (D06) and Kristian Björk (D05)

Abstract

Embedded sound positioning software

Kristian Björk (D06) and Kristian Björk (D06)
Start:   2010-09-01

The surveillance intelligence in cameras is increasing, for the most part
it is video applications. This thesis is a push forward for embedded audio
intelligence. On the market today there are similar high end dedicated
sound detection systems. To be able to use large networks of cameras to
detect the position of sound is a difficult task and requires a lot of the
camera. This thesis is to be a proof of concept.

The goal of this master's thesis is to demonstrate that it is possible to
construct a sound positioning system under simple circumstances. This task
is both theoretical and practical. Investigate different ways to identify
the sound and determine ways to improve the performance. All is to be
implemented in an Axis network camera. Suggest improvements for audio in
current Axis software. Evaluate future improvements and possibilities to
the prototype.

    * Determine the possible precision of off position estimates
    * Suggest methods for sound identification
    * Implement a proof of concept on an Axis network camera
    * Cancellations of unwanted sound artifacts such as echo


Advisor: Bengt Mandersson (EIT)
Advisor: Per Vilhelmsson  (Axis Communications AB)