Algorithm to estimate Weight from walk over scale

Estimate weight from measurements from the sample data set. The data set is very dirty because one asset can step in the platform while other asset is steeping out. each asset is identified by a number. Run the python code ([login to view URL]) to see a live demo of the data. The example code uses an exponential moving mean to calculate changes in the direction of the data(peaks and valleys) and calculate the distance of the previous peak/valley to identify meaningful change. The code will be run in an embedded system in "real-time"(there's a delay because the moving average), just keep enough data to do the segmentation, then, clear the memory. The provided code is just an idea... you can be creative. Requirements: C++ implementation of a class with update method like the example in python, with other methods to get the weight/asset_id The algo must return usable data points at least 20% of the times, that means, we get one valid reading out of four walks in a crowded situation. For now just use the mean to estimate weight, no fancy algos, no machine learning. The algo must consider when asset walks alone in the platform(the easiest part)