Hello everyone, I'm new in this filed please help!.
The problem that I'm working on is ECG signals classification using recurrent neural network (RNN).
I attempt to use one of the following Matlab toolboxes for building the RNN:
https://github.com/yechengxi/LightNet.
http://pyrenn.readthedocs.io/en/latest/index.html.
The dataset that I'm working on consists some ECG records. Since RNN can be training using high-level features, at first step I found the R peaks for each ECG signal, which resulting in one vector for each ECG signal, then I combine all these vectors in one matrix (20*5000), where 20 corresponding to rows (number of ECG records), and 5000 corresponding to columns (number of maximum sample of ECG record) because the number of samples is different for each ECG signal.
This matrix has to be used for training the RNN. Since this process is supervised learning, I should create the matrix for the Labels (targets), I'm a little bit confused how to create such labels. For example, if we have two classes (class 1: representing 0, and class 2: representing 1 for each ECG record), do we should have the matrix of 20*5000 of the same size as the training data, where (20 labels= rows: number of ECG record) and (5000 samples= columns: represent 0 or 1 ). Please help me if you have any information that can help.
This is a good queston.
Examples of how to prepare training data for a RNN are given in
https://www.researchgate.net/publication/3302435_Recurrent_neural-network_training_by_a_learning_automaton_approach_for_trajectory_learning_and_control_system_design
and
https://www.researchgate.net/publication/2740004_Extended_Kalman_filter_in_recurrent_neural_network_training_and_pruning
More to the point, see
https://www.researchgate.net/publication/220220141_Recurrent_neural_networks_employing_Lyapunov_exponents_for_analysis_of_ECG_signals
Article Recurrent neural-network training by a learning automaton ap...
Article Extended Kalman filter in recurrent neural network training ...
Article Recurrent neural networks employing Lyapunov exponents for a...
You write your mathematical equation for ECG signals and range/value of the different parameter and constant in the equation and create a .mat file, use this data (70%+30%) for training and testing.
Dear Huda
I suggest using Matlab simulink.
Regards
- Badges
- Science topic