Updated 10 May 2016
These files contain all the code necessary to run the example in the Webinar "Signal Processing and Machine Learning Techniques for sensor Data Analytics". They also include code to automate the download and preparation of the dataset used.
In that webinar (http://www.mathworks.com/videos/signal-processing-and-machine-learning-techniques-for-sensor-data-analytics-107549.html) we presented an example of a classification system able to identify the physical activity that a human subject is engaged in, based on the accelerometer signals generated by his or her smartphone. We discussed signal processing methods to extract highly-descriptive features, and we gave an overview of a number of techniques to choose and train a classification algorithm. Along the way we demonstrated the use of Parallel Computing to accelerated the extraction of features from a large dataset.We also presented a workflow to transition signal processing and predictive algorithms to embeddable software implementations - first using DSP system modelling, and then automatically generating C/C++ source code directly from MATLAB.
Excelent, but I've to point out a minor issue during execution of "extractALLFeatures.m".
For k = 2714 it simply crashes. I was looking the code and noticed that the problem is on the autocorrFeatures() helper function (on "extractSignalFeatures.m").
At line 69, it initializes feats as zero(1,3). But at the line 91, if it's only able to find one peak (therefore, f1 only) it assumes that f2 and f3 are already 0.
So in the case of k = 2714, when extracting autocorrFeatures for atz, it crashes, because it expects a 1 by 9 matrix, but it returns a 1 by 7 matrix.
Thank you for rating the submission!
The webinar already includes a link to this item from the bottom of the page, so I am conscious I may not have well understood your suggestion. Would you mind if I checked this with you directly? You can reach me at gabriele <dot> bunkheila <at> mathworks <dot> com.
Maybe you should add a link to the webinar of Huard or link the files on the webinar page. They were not easy to identify.
But the material is great. Thank you!
Added hyperlink to webinar page