Laser & Optoelectronics Progress, Volume. 58, Issue 16, 1600006(2021)
Functional Near-Infrared Spectroscopy-Based Brain-Computer Interface
Figures & Tables(10)
Fig. 1. Quantity of literatures related to fNIRS-BCI recorded in the Superstar Discovery database over the years
Fig. 4. Proportion of main feature classification algorithms used in fNIRS-BCI from 2004 to 2020
Fig. 5. Proportion of main feature classification algorithm used in fNIRS-BCI from 2004 to 2014 (data come from Ref.[10])
Table 1. A list of the quantity of literatures related to fNIRS-BCI recorded in several databases
View tableTable 1. A list of the quantity of literatures related to fNIRS-BCI recorded in several databases
Database Superstar Discovery Web of Science Engineering Village China National Knowledge Infrastructure Quantity of entries 478 217 165 70
Table 2. Retrieval results of feature classification algorithm used in fNIRS-BCI from 2004 to 2020 (Based on Superstar Discovery database; the statistical period is until June 2020)
View tableTable 2. Retrieval results of feature classification algorithm used in fNIRS-BCI from 2004 to 2020 (Based on Superstar Discovery database; the statistical period is until June 2020)
Retrieval content Subject Abstract All the fields Average SVM/Support Vector Machine 67/96 61/89 67/97 79.50 LDA/Linear Discriminant Analysis 40/70 38/68 40/70 54.33 ANN/Artificial Neural Network 10/19 9/15 10/22 14.17 DNN/Deep Neural Network 7/19 7/19 7/19 13.00 CNN/Convolutional Neural Network 3/8 3/7 3/8 5.33 HMM/Hidden Markov Model 3/4 3/4 3/4 3.50 LR/Logistic Regression 3/4 3/4 3/4 3.50 QDA/Quadratic Discriminant Analysis 3/3 3/3 3/3 3.00 GBDT/Gradient Boosting Decision Tree 3/3 3/3 3/3 3.00 NB/Naive Bayes 4/2 4/2 4/2 3.00 RF/Random Forest 2/2 2/2 2/2 2.00 KNN/K-Nearest Neighbor 1/1 1/1 1/1 1.00
Table 3. Quantity of fNIRS-BCI literatures related to SVM and LDA included in several database from 2015 to 2020
View tableTable 3. Quantity of fNIRS-BCI literatures related to SVM and LDA included in several database from 2015 to 2020
Classification algorithm Superstar Discovery Web of Science Engineering Village Average SVM 54 34 22 37 LDA 31 23 14 23
Table 4. Summary of major fNIRS-BCI studies from 2015 to 2020
View tableTable 4. Summary of major fNIRS-BCI studies from 2015 to 2020
Reference Task Feature Classification algorithm Number of classification Classification accuracy Ref.[93], 2015 MA and other 11 mental tasks Slope of HbO, HbR and HbT, etc. LDA 2 77%(personalized tasks); 73%(prescribed tasks) Ref.[94], 2015 MA and mental counting Mean of HbO and HbR SVM 2 82.4%(the best) Ref.[95], 2015 MA, word generation, mental rotation, and rest Slope of HbO and HbR DNN, LDA, and SLDA 2 DNN: 63.2%; LDA: 64.3%; SLDA: 65.7% Ref.[96], 2015 Active and passive movements of fingers of the right hand Average level of HbO and spatial patterns SVM 2 71.82% Ref.[97], 2017 MI and ME(hand extension and finger tapping) Mean, slope, quadratic coefficient, and approximate entropy of HbO SVM 4 87.65%(hand extension); 87.58%(finger tapping) Ref.[82], 2016 MA and rest Mean, peak, slope, variance, kurtosis, and skewness of HbO LDA, QDA, KNN, NBC, SVM, and ANN 2 71.6%, 90.0%, 69.7%, 89.8%, 89.5%, and 91.4%(2D features); 79.6%, 95.2%, 64.5%, 94.8%, 95.2%, and 96.3%(3D features) Ref.[98], 2017 ME of left and right hand(hand grasping) Approximation coefficients of EEG and mean of HbO and HbR SVM 2 91.02% Ref.[99], 2017 Visual task of primary RGB colors Mean, peak, slope, skewness, and kurtosis of HbO LDA 3 55.29% Ref.[71], 2017 MI of left and right hand Mean,slope, and variance of HbO and HbR LDA 2 75.3%(HbO); 72.3%(HbR) Ref.[100], 2017 Handshake, ball grasp, poking, and cold temperature stimulation Mean, peak, and skewness of HbO LDA 2 and 4 Binary: 78.12%, 75.94%; quaternary: 50.31% Ref.[87], 2017 ME of left and right hand and rest Mean, peak,slope, variance, kurtosis, and skewness of HbO and HbR SVM, ANN, and CNN 3 SVM: 86.19%; ANN: 89.35%; CNN: 92.68% Ref.[101], 2018 ME of finger tapping Mean and slope of HbO and HbR LDA and GLM 2 LDA: 78.7%; GLM: 65.76%(HbO), 70.3%(HbR) Reference Task Feature Classification algorithm Number of classification Classification accuracy Ref.[72], 2018 MA and rest Mean and variance of HbO and HbR LDA 2 77.2%(HbO); 72.9%(HbR) Ref.[77], 2018 Stroop task Wavelet energy of HbO SVM 2 87.31%(6 channels); 74.31%(16 channels) Ref.[57], 2019 Speech imagery Mean of HbO RLDA 3 64.1% Ref.[102], 2019 Drowsy and alert states of driving a car simulator Automatically extracted from different channels and time windows DNN and CNN 2 DNN: 97.2%(the best); CNN: 99.3% Ref.[60], 2020 MI, word generation, n-back and discrimination/selection response task Automatically extracted from EEG, HbO, HbR, and HbD DNN and SVM 2 and 4 MI task:91%(DNN), 85%(SVM); n-back task: 87%(DNN), 82%(SVM) Ref.[103], 2020 Emotional task Average level of HbO and HbR LDA 2 64.50%, 67.75%, 71.00% Ref.[92], 2020 Self-paced walking TKE and AUC of HbO, HbR, and HbT GBDT, RF, LR, and LDA 2 AUC:94.4%, 93.6%, 82.6%, and 81.9%; TKE(GBDT): 98.75%(offline), 100%(pseudo-online) Ref.[104], 2020 MI of left and right hand, MA and rest Mean, slope, peak, skewness, and kurtosis of HbO and HbR LDA, KNN, and SVM 2 LDA: 87.87%; KNN: 79.59%; SVM: 89.54%(the best)
Table 5. Comparison of various models of BCI
View tableTable 5. Comparison of various models of BCI
BCI mode Spatial resolution Temporal resolution Resistance to motion disturbance Resistance to electromagnetic interference Portability Equipment cost Novelty EEG * *** * * *** * * PET *** * * NA * *** * fMRI *** * * NA * *** * fNIRS ** ** ** *** *** ** ***
Tools
Get Citation
Copy Citation Text
Hongyun Li, Yunfa Fu. Functional Near-Infrared Spectroscopy-Based Brain-Computer Interface[J]. Laser & Optoelectronics Progress, 2021, 58(16): 1600006
Paper Information
Category: Reviews
Received: Sep. 8, 2020
Accepted: Oct. 21, 2020
Published Online: Aug. 16, 2021
The Author Email: Fu Yunfa (fyf@ynu.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20