The world around us is filled with a vast array of sounds. Many of these sounds do not conform to the relatively predictable patterns present in speech and in music, and significant sounds such as a smoke alarm, breaking glass or a gunshot are not easily distinguishable by a computer from the background hum and bustle of everyday life.
Audio Analytic was founded in 2010 by Dr Chris Mitchell to commercialise his PhD research in this area. Audio Analytic’s software uses the technology of machine learning to make sense of environmental sounds, without needing to resort to huge computational power. Its system is divided into a feature extraction module which extracts the particular characteristics of sounds as the ears do, and a pattern matching engine which learns from data as the brain does.
The software can be plugged into relatively simple smart home devices such as video cameras, lights and plug hubs, providing a wide range of potential applications within the home. It can then send you an alert via smartphone or tablet if, for example, it detects the sound of breaking glass in your home or a smoke alarm going off. It can allow vulnerable users to summon help without being within reach of a panic button, or work as a baby monitor.
Audio Analytic already has a number of customers including Swann, Cisco and Zenitel.