I'm working with an accelerometer-based dataset capturing accelerations from the human thigh. Most HAR processes will have window segmentation to break down the signal into samples, and pre-processing of the signals (i.e filtering) prior to feature extraction. By doing both processes on the same signal from my dataset and subtracting the difference, I've found that the resulting output signal is different depending on the order these two operations are carried out.
Which of these processes should be carried out first? In the literature, I have seen several instances of either approach being taken.
Article Recent trends in machine learning for human activity recogni...
I had to perform segmentation of human speech for the purpose of my PhD. In my experience, you're better off to keep your original signal/data, and segment it prior to other processing : your decision will be cleaner, and reversible at the same time.
My own real time segmentation was based on the detection of biological events (glottal closure) comparing two sliding models, one short, one long, and using a cumulated sum of Kullback-Leibler divergence estimates, looking for convexity jumps in this criterion (D2f(t) = f(t+1) - 2f(t) + f(t-1) estimates the second order derivative of signal f...).
Get to know the semantics of what you are looking for, the the models will fall in place smoothly
Hope this helps
Ref: my IEEE Journal article on real time speech segmentation and coding, posted on researchgate
Because Neural networks use non-linear activation functions these systems are not-linear. In this regard "the resulting output signal is different depending on the order ... operations are carried out." In any case, you need to use experimentation for better choices.
Dear Mr. Jamieson!
You raised an important issue. I take COVID-19 as a case-study - context:
Shorten, C., Khoshgoftaar, T.M. & Furht, B. Deep Learning applications for COVID-19. J Big Data 8, 18 (2021). https://doi.org/10.1186/s40537-020-00392-9, Citation: "Our survey provides a description of how some of these problems can be solved with Deep Learning. We have described how different data types are inputted to Deep Neural Networks and how tasks are constructed as learning problems....", Open access:
Article Deep Learning applications for COVID-19
Based on my interpretation, it is better first to segment and after that filter datasets.
Yours sincerely, Bulcsu Szekely
- Badges
- Science topic