Visual Analysis of Acceleration Sensor Data
This is a sample F# app that uses Xamarin and OpenCV to perform 'visual' processing of linear acceleration sensor data.
The main idea is to take a sample of sensor data and convert that into a picture in delay coordinates. Essentially we are freezing the window of sensor information (which could be 100's of elements) into a single composite picture.
The picture can then be processed using a Hough transform to extract structured information in the form of line equations - in polar coordinate - for the dominant lines discovered in the picture.
The line equations are a compression of the sensor data to a low dimensional space - you can choose to use the top n lines or pick all lines over some threshold. This compressed view can then be used for further processing such as classification or clustering, etc.
Screen shot:
