CapoCaccia

Cognitive Neuromorphic Engineering Workshop

User Tools

Site Tools


cc18:exploring-the-classification-of-biomedical-signals-using-plasticity-in-recurrent-networks-of:overview

hello! this is the biomedical signals processing wiki…

The idea of the workgroup is to implement a liquid state machine and classify the muscular movement form EMG upper-limb recordings. The platform we are going to use is Brian2 (http://briansimulator.org).

BRIAN2 Here is an example of LSM in Brian: https://github.com/ricardodeazambuja/SNN-Experiments/blob/master/Implementing%20a%20Liquid%20State%20Machine%20using%20Brian%20Simulator/Implementing%20a%20Liquid%20State%20Machine%20using%20Brian%20Simulator.ipynb

To install Brian2: http://brian2.readthedocs.io/en/stable/introduction/install.html

To write the cose we suggest to use Jupiter Notebook, it is a web-application: http://jupyter.org/index.html

Once is installed to open it you should write: Jupiter notebook in the terminal. If you have problem with Mac Sierra OS, try: python -m IPython notebook

State of the art… Article about an example of recurrent neural network ap_stdp_lsm.pdf

Practical example of RNN implementation practicalesn.pdf

Liquid state machine article (Maass, Markram) lsm.pdf

Example of LSM application with SNN corrai.pdf

On-line papers: AP-STDP: A novel self-organizing mechanism for efficient reservoir computing (Yingyezhe Jin & Peng Li)

Spiking neural networks, an introduction (Jilles Vreeken)

A Practical Guide to Applying Echo State Networks (Mantas Lukosevicius)

sample data set of finger motions nengo_training.csv.zip

subgroups and challenges (please, if you know how to rotate the pictures…do it…) THANKS!

DATA SET so guys, here is an amazing data set of people playing pseudo random rock-paper-scissors game. have fun! the format: myo_to_csv.write(time.time(), myo.emg, myo.quat, myo.acc, myo.gyro, label) data.zip

code to extract timestamp/emg/label In the folder there are: folder with recordings with subject name, folder with the recordings with file without subject name. parsing_data.py is a file that extract the emg signals: G is the output: the first column is the timestamp and the other 8 are the egg signals. The labels are saved in the vector annotation. In this file you should pass the path of a single file. the last folder is for the spikily, emg2spikes.py is the main code. In level crossing,py there is the function for loading the data and to create spikes. parsing_data_and_spikify2.zip data_spikes.zip Folder data_spikes contains a folder for each subject and one for each muscle with inside the up and down channel and the list of annotation.

Sinewave and spikes Here the code to produce spikes from a sine wave. spikes_sinewave4.py.zip

Paper The real Bucket paper: https://pdfs.semanticscholar.org/af34/2af4d0e674aef3bced5fd90875c6f2e04abc.pdf

Matlab Hey here is the Matlab code I wrote so far to have look at the data and train an svm. It works now for data of Melika. It works less good for Valerio. The polar plot kind of gives it away. (M)

emg_svm.zip

New Rock data: rock2.zip

New Rock data: scissor5.zip

Ok; All data needed to make the Myo band Demo - “ play against Elisa ”-) working myo_capocacciafiles.zip

cc18/exploring-the-classification-of-biomedical-signals-using-plasticity-in-recurrent-networks-of/overview.txt · Last modified: 2019/05/16 20:20 (external edit)