Latent common source extraction (LCSE)
LCSE seeks to construct a common latent representation of the SSVEP signal subspace that is stable across multiple trials of EEG data.
The spatial filter thus obtained improves the signal-to-noise ratio (SNR) of the SSVEP components by removing nuisance signals that are irrelevant to the generalized signal representation learnt from the given data.
This is a sample code for testing LCSE as described in the article [1].
Note: Use the SSVEP benchmark dataset placed in a folder in the same directory as this file
Link: ftp://sccn.ucsd.edu/pub/ssvep_benchmark_dataset/
Dataset used: SSVEP benchmark dataset [2]
A 40-target SSVEP benchmark dataset recorded from 35 subject. The stimuli were generated by the joint frequency-phase modulation (JFPM) [1]
-
Stimulus frequencies : 8.0 - 15.8 Hz with an interval of 0.2 Hz
-
Stimulus phases : 0pi, 0.5pi, 1.0pi, and 1.5pi
-
No. of channels : 64
-
Selected channels :9 electrodes[ 1(48): Pz, 2(54): PO5, 3(55): PO3, 4(56): POz, 5(57): PO4, 6(58): PO6, 7(61): O1, 8(62): Oz,and 9(63): O2)
-
No. of recording blocks : 6
-
Data length of epochs : 6 s [ 0.5s cue, 5s data, 0.5s blank ]
-
Sampling rate : 250 [Hz]
Data is a 4-D matrix named "data" with dimensions of [64, 1500, 40, 6] The four dimensions indicate, [Electrode_index, Time_points, Target_index, Block_index]
Required functions: LCSE
The function has been test only on the SSVEP benchmark dataset described in [2]
[prediction, score] = LCSE(sampling_rate, train_data, test_data, number_of_filterbank, number_of_Reconstructed_channel);
Filter bank coefficients : a = 1.25; b = 0.25; as defined in [1]
Specification of each input variables
-> sampling rate - EEG data sampling rate (integer value)
-> train_data - data dimension -> [no. of channels, no. of EEG data points, no. of targets, no. of trial blocks]
constraints: no. of channels >=2
no. of EEG data points >= 0.2s of data points
no. of targets >=2
no. of trial blocks >=2
-> test_data - data dimension -> [no. of channels, no. of EEG data points,no. of targets]
constraints: the dimensions of test data should match the corresponding
Dimension in the train data
-> number_of_filterbank - should be a interger value between 1 and 7, Refer [1] for further description
-> number_of_Reconstructed_channel - should be a interger value between 1 and no. of trial blocks used for training, Refer [1] for further description
Reference: [1] Kiran Kumar G. R and Ramasubbareddy M, "Latent common source extraction via a generalized canonical correlation framework for frequency recognition in SSVEP based brain–computer interfaces," in Journal of Neural Engineering. doi: https://doi.org/10.1088/1741-2552/ab13d1
Visit my research gate page to download the preprint, link : https://www.researchgate.net/profile/Kiran_Kumar_G_R
[2] Y. Wang, X. Chen, X. Gao and S. Gao, "A Benchmark Dataset for SSVEP-Based Brain–Computer Interfaces", in IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 25, no. 10, pp. 1746-1752, Oct. 2017.
Author: Kiran Kumar G R
Affiliation: Indian Institute of Technology Madras.
email: kirankumar.g.r@hotmail.com
Google Scholar: https://scholar.google.co.in/citations?user=fH96otoAAAAJ&hl=en
Last revised : November 2019
Please send suggestions/queries to the above email