This is an implementation of an anomaly detector based on subspace projection methods. It looks for changes in sounds detected by a microphone. You first train it on the "normal" sound of e.g. a motor, then you let it run a loop while it listens for deviations in the received sound from the training sound.
The code runs on a BeagleBone, with an ADC-001 A/D cape, which is a 2-channel A/D board designed for data acquisition in the audio range (50Hz -- 15kHz). The cape takes input from a MIC-001 microphone. Complete documentation for the ADC-001 cape and MIC-001 microphone is available under:
https://github.com/brorson/ADC-001_hardware_information
These hardware items are available for sale at BNM-Hobbies:
http://www.bnm-hobbies.com/store/index.php?main_page=index&cPath=150
Stuart Brorson March, 2018.