Algorithm overview

Every time a new frame from the camera is available, the difference image between the new frame and the previous frame is calculated. When 15 difference images (every 0.5 s with 30 fps) has been calculated, an asynchronous task is started to estimate the compression rate, if compression is detected. An asynchronous task is a task that runs in the background. Here it is used to avoid lag in the user interface and to prevent frame drops. This task has the biggest workload of the algorithm. It uses the 15 new difference images provided, as well as up to 75 of the previously used (making it a total of up to 90 images, equal to 3 s with 30 fps). First, the task tries to establish an ROI. If an ROI can not be established yet, the whole frame is used as ROI. The sum of changes is calculated inside the ROI, using up to 90 of the previous images. We use FFT and some other techniques to estimate the dominating frequency in the window we are examining. This dominating frequency has to pass several tests before being trusted. If the frequency is not trusted it means the algorithm does not believe that it is caused by chest compressions, but by some other movement. If trusted, a weighted moving average filter is applied to the frequency, and thereafter it is transmitted and displayed as compression rate in cpm.