Skip to main content
. 2021 Jun 23;21(13):4293. doi: 10.3390/s21134293

Table 3.

Summary of related works on embedded BCIs for functional substitution: N/I: No indication, 0: bad, 1: good.

Work Year Cs Ad Algorithms Accuracy (%) Platform Time (ms) Power (W) Online/ Offline
[13] 2008 0 0 Average filter, PCA, Linear regression 74.6 Virtex 6 2420 1.11 Online
[49] 2010 1 0 FIR, DWT, SVM N/I Compact-RIO N/I 12.81 Offline
[51] 2015 1 0 Theta spectra, threshold 71 FPGA Mobile tablet N/I 4 Online
[52] 2010 1 0 STFFT, ICA, threshold 78.24 DSP N/I 1.11 Online
[53] 2017 0 0 FFT, threshold 99.4 Micro2440SDK ∼8500 24 Online
[54] 2013 1 0 IIR, DWT, threshold hierarchical model 91 CompactDAQ 2 12 Online
[55] 2016 0 1 DWT, SVM 82.1 Odroid-xu4 0.11 20 Offline
[56] 2017 0 0 Band-pass filter, average power, temporal correlation 70 Arduino Due MCU 2.23 1 Offline
[57] 2014 0 0 FFT, SLIC 70 Micro2440 (ARM) 0.1 24 Offline
[58] 2017 0 1 ERD/ERS, Adaptive Threshold 75 Zynq ZC7030 402 4.12 Online, offline
[59] 2014 0 1 FFT, Mahalanobis distance 82 Blackfin, DSP N/I 4.02 Online
[60] 2014 0 0 FFT 98.8 AT89S51, Tablet (AUSU) 42 12 Online
[48] 2010 0 0 Phase coding, FFT 89.29 Cyclone EP2C20Q FPG 30.14 ∼27 Online
[61] 2012 1 0 FFT, Mardia test, Mahalanobis distance (MD) 77.6 Cyclone EP2C20Q FPG 30.14 ∼27 Online
[14] 2012 1 0 Threshold 61.6 iPhone 32 ∼6 Online
[62] 2012 0 0 FFT, Morlet Continuous Wavelet, Threshold N/I Spartan3 XC3S1400AN 1 N/I Online
[63] 2011 0 0 PSD, LDA 73.58 ASIC, MSP430F1611, NRF24L 200 0.001395 Online
[23] 2017 1 1 Adaptive filter, CSP, MD 94.47 Stratix-IV 394 1.067 Offline
[64] 2018 1 1 WOLA filter bank, CSP, MD 80.2 Stratix-IV 430 0.67 Online, Offline
[65] 2012 0 0 Forward Filter, FLDA 73.96 Spartan 3E FPGA N/I N/I Online
[66] 2012 0 0 FIR filterbank, Hidden Markov Models 76.5 Spartan 6 FPGA N/I N/I Online
[67] 2016 1 0 adaptive filtering N/I FPGA Virtex-5 LX50T N/I N/I Online
[68] 2020 0 0 PSD, band-pass filtering, canonical correlation analysis (CCA) 80 XC7K325T-2FFG900C N/I N/I Online
[69] 2017 0 0 Median, FIR filter, FLDA N/I Virtex-5 0.01 0.67 Online
[70] 2015 0 0 ICA, Canonical Correlation Analysis (CCA) 86.5 FPGA 32000 N/I Online
[71] 2016 0 0 Blind Source Separation (BSS), CCA 93.41 FPGA N/I N/I Online
[72] 2018 0 0 Sparse Bayesian Learning (BSBL), multi-layer perceptron regressor 89.85 Virtex7, ARM N/I N/I Online
[47] 2019 0 1 CNN 80.5 Xilinx BNN-PYNQ 1.97 0.025 Offline
[73] 2015 0 0 FFT, Threshold 88.88 Xilinx & PC Tablet 4430 70 Online
[74] 2018 1 0 Surface Laplacian, Separable Common Spatio Spectral Pattern (SCSSP), Mutual Information (MI), Linear Discriminant Analysis (LDA), and Support Vector Machine (SVM) 81.9 Virtex-6 FPGA 0.550 83 Online & offline
[75] 2020 0 0 FIR filter, averaging method 90.62 Cyclone II EP2C35 DSP ∼2000 ∼27 Online
[76] 2019 1 0 Channel selection, wavelet, energy normalization, LDA 80 Xilinx 7.5 0.102 Offline
[77] 2013 0 0 FFT, threshold 92.5 FPGA, MCP3201 microcontroller 5200 1.74 Online
[78] 2020 0 0 Canonical correlation analysis (CCA) 76 DE0-nano board 0.00052 ∼0.05 Online
[79] 2019 0 0 CCA 87.89 Cyclone IV EP4CE115 1500 ∼0.05 Offline
[80] 2020 0 0 Long Short-Term Memory (LSTM) 87.89 MindReading photonic ULQ 1500 0.2155 Offline
[81] 2019 0 0 bandpass Butterworth filter, DWT, Feedforward Neural Network (FFNN) 96.09 Raspberry Pi 3B N/I 5.77 Online