Abstract:Currently, steady-state visual evoked potentials (SSVEPs) are applied in a variety of fields. In these applications, spatial filtering is the most commonly used method for decoding SSVEPs. However, existing methods for the SSVEP decoding require improvements in terms of the interaction speed, accuracy, and ease of use. Recent advances in performance have been attributed to the existing hybridisation-based approaches. This article presents and evaluates a novel approach for target identification of SSVEP. The suggested target recognition method significantly enhances the performance of SSVEP decoding by combining two existing methods: extended canonical correlation analysis (eCCA) and the sum of squared correlation (SSCOR). This combined approach is referred to as the extended CCA and sum of squared correlation approach (eCCA-SSCOR). The results of the offline experimental comparison are based on two publicly available datasets, including the San Diego dataset, which comprises 12 frequency targets recorded from 10 individuals, and the Benchmark dataset, comprising 40 frequency targets from 35 subjects. The statistical significance of the improvement was tested using paired samples t -tests and Wilcoxon rank-sum tests. The results indicate that the suggested eCCA-SSCOR approach significantly improves detection accuracy and ITR compared to four existing state-of-the-art methods: canonical correlation analysis (CCA), filter bank CCA (FBCCA), eCCA, and SSCOR. Our method achieved a high average accuracy for all subjects, with 99.5% for the San Diego dataset, and 98.60% for the Benchmark dataset corresponding to data length of 3.5 s and 5 s, respectively. Furthermore, a high ITR of 212.44 bits/min was achieved with a data duration ( ) of 0.75 s using the San Diego dataset and was 238.66 bits/min with a of 1 s for the Benchmark dataset. These offline results demonstrate that the proposed approach successfully combines eCCA and SSCOR to improve SSVEP decoding. Additionally, it is advantageous for applications based on offline processing and can be adapted for online analysis. Online applications can benefit from these results by providing fast interfaces with a large number of commands.

SSVEP Harmonic Fusion for Improved Visual Field Reconstruction with CNN

Research on SSVEPs in Correlational Convolutional Neural Networks Based on Filter Banks

MHLCNN: Multi-Harmonic Linkage CNN Model for SSVEP and SSMVEP Signal Classification

Multiscale Noise Suppression and Feature Frequency Extraction in SSVEP Based on Underdamped Second-Order Stochastic Resonance

An SSVEP Classification Method Based on a Convolutional Neural Network

Compact Convolutional Neural Networks for Classification of Asynchronous Steady-state Visual Evoked Potentials

SSVEP-EEG Feature Enhancement Method Using an Image Sharpening Filter

Convolutional Correlation Analysis for Enhancing the Performance of SSVEP-Based Brain-Computer Interface

SSVEP-EEG Denoising Via Image Filtering Methods.

Steady-State Visual Evoked Potential Classification Using Complex Valued Convolutional Neural Networks

Multi-scale noise transfer and feature frequency detection in SSVEP based on FitzHugh-Nagumo neuron system

High-Order Temporal Convolutional Network for Improving Classification Performance of SSVEP-EEG

OS-SSVEP: One-shot SSVEP classification

A Deep Learning Method for SSVEP Classification Based on Phase and Frequency Characteristics

A Time-Local Transformation Steady-State Visual Evoked Potentials Recognition Framework Combined With Filter Banks

A Convolutional Neural Network for the Detection of Asynchronous Steady State Motion Visual Evoked Potential

An efficient CNN-LSTM Network with spectral normalization and label smoothing technologies for SSVEP frequency recognition

Approach to Recognizing SSVEP by Combining CCA with Artificial Neural Network

Hybrid approach: combining eCCA and SSCOR for enhancing SSVEP decoding

Capturing Pictures from Human Vision Using SSVEP and Lock-in Amplifier

Decoding SSVEP Via Calibration-Free TFA-Net: A Novel Network Using Time-Frequency Features