Activity Recognition Using Sound
As technology and internet rapidly increases, software, and automation, smart home technology products make everyone’s life easier and more secure. UbiSound Home Monitor is an Wifi-enabled smart home secure device to help you monitor your door no matter where you are or what you're doing. When a visitor knock the door or close the door, you will be notified by a text message.
How it Works
By ‘listening’ to home sound waves in a wide frequency ranges, combined with machine learning, the device is able to monitor several activities, including closing door and knocking door. This project involves implementing hardware platform, collecting sound waves, building sound classifier, real-time signal processing, and real-time message broadcasting. Sound detection is performed by analyzing the short audio file recorded every 5 seconds extracting several signal processing techniques, including short-term energy, zero-crossing rate, and the sum of a range of frequencies. The detection algorithm utilized fast Fourier transform from numpy library, and machine learning from AWS.
Existing Work
Background sound has also been used for human activity recognition in numerous applications. Such systems that can accurately identify human activities from sensor data might provide activity specific assistance or implicitly create activity logs for wellness and secure monitoring. One of the famous recognition method is Mel Frequency Cepstral Coefficients (MFCC) and Dynamic Time Warping (DTW). They can produce accuracies for a variety of human activity, including . For activity like watching television. Sound could be a much better predictor than physical motion in many indoor activities.
There are already a number of existing product on the market. One is Leeo Smart Alert. It works with the smoke and carbon monoxide detectors that you already have to notify you when something happened. It is designed to only listen to the sound of smoke and CO alarms. If any smoke or CO detector goes off, Leeo calls you instantly on your phone and tells you what’s happening. Such product is very similar to our product, except we are listening to something else. We believe this is a rapidly growing market as the increasing demand on smart-home devices. |