Approved
Gunshot Detection from Audio Streams in Portable Devices
Ellen Grane () and Linnea Bokelund ()
Start
2022-01-16
Presentation
2022-06-09 10:15
Location:
E:3139
Finished:
2022-06-17
Master's thesis:
Abstract
In society today there exists a multitude of devices able to capture audio. The cellphone, which almost everyone today constantly carry in their pocket, is one example of such a device. For cellphones, video cameras and other similar devices there would for some situations be beneficial if the audio could be used to automatically predict if a situation is threatening or not. This information could then be used to trigger actions to protect, alert etc. For example, while walking home at night the cellphone could automatically call an emergency number if a threatening situation can accurately be detected. One part that are interested in a feature like this is the Axis Body Worn Solutions. They develop wireless body mounted video cameras used by police officers, guards and more. Currently the user starts a recording manually by pressing a button, alternatively a recording can be triggered if the user falls over. We are interested to see if it is possible to detect threatening situations from audio, and we will examine this by using the audio from Axis Body Worn cameras, and based on that trigger recordings. As portable devices usually have a limited battery, memory and processing power we also want to examine the effect that such detection has on the devices’ energy, CPU and memory consumption.
Supervisor: Pierre Nugues (LTH/CS)
Examiner: Maria Kihl (EIT)