Impact of different signal durations on absolute error.

<p>Data sets with different signal durations were analysed using all the methods and the average absolute error is presented for each test signal shapes. Data sets were created by adding 160% walking noise to the hourly-sampled templates of different duration. A) cosine data, B) pulse data, C) double pulse data, D) DNFL shoulder data, E) DNFL asymmetry data, F) aggregated results from all the shapes.