Improving Generalized Zero-Shot Learning SSVEP Classification Performance From Data-Efficient Perspective
Generalized zero-shot learning (GZSL) has significantly reduced the training requirements for steady-state visual evoked potential (SSVEP) based brain-computer interfaces (BCIs). Traditional methods require complete class data sets for training, but GZSL allows for only partial class data sets, divi...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2023-01-01
|
| Series: | IEEE Transactions on Neural Systems and Rehabilitation Engineering |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10283906/ |
| _version_ | 1797650815985385472 |
|---|---|
| author | Xietian Wang Aiping Liu Le Wu Ling Guan Xun Chen |
| author_facet | Xietian Wang Aiping Liu Le Wu Ling Guan Xun Chen |
| author_sort | Xietian Wang |
| collection | DOAJ |
| description | Generalized zero-shot learning (GZSL) has significantly reduced the training requirements for steady-state visual evoked potential (SSVEP) based brain-computer interfaces (BCIs). Traditional methods require complete class data sets for training, but GZSL allows for only partial class data sets, dividing them into ‘seen’ (those with training data) and ‘unseen’ classes (those without training data). However, inefficient utilization of SSVEP data limits the accuracy and information transfer rate (ITR) of existing GZSL methods. To this end, we proposed a framework for more effective utilization of SSVEP data at three systematically combined levels: data acquisition, feature extraction, and decision-making. First, prevalent SSVEP-based BCIs overlook the inter-subject variance in visual latency and employ fixed sampling starting time (SST). We introduced a dynamic sampling starting time (DSST) strategy at the data acquisition level. This strategy uses the classification results on the validation set to find the optimal sampling starting time (OSST) for each subject. In addition, we developed a Transformer structure to capture the global information of input data for compensating the small receptive field of existing networks. The global receptive fields of the Transformer can adequately process the information from longer input sequences. For the decision-making level, we designed a classifier selection strategy that can automatically select the optimal classifier for the seen and unseen classes, respectively. We also proposed a training procedure to make the above solutions in conjunction with each other. Our method was validated on three public datasets and outperformed the state-of-the-art (SOTA) methods. Crucially, we also outperformed the representative methods that require training data for all classes. |
| first_indexed | 2024-03-11T16:07:05Z |
| format | Article |
| id | doaj.art-421d8cd095d948a8ac0d71dac8787eaf |
| institution | Directory Open Access Journal |
| issn | 1558-0210 |
| language | English |
| last_indexed | 2024-03-11T16:07:05Z |
| publishDate | 2023-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Transactions on Neural Systems and Rehabilitation Engineering |
| spelling | doaj.art-421d8cd095d948a8ac0d71dac8787eaf2023-10-24T23:00:09ZengIEEEIEEE Transactions on Neural Systems and Rehabilitation Engineering1558-02102023-01-01314135414510.1109/TNSRE.2023.332414810283906Improving Generalized Zero-Shot Learning SSVEP Classification Performance From Data-Efficient PerspectiveXietian Wang0https://orcid.org/0000-0002-0571-8658Aiping Liu1https://orcid.org/0000-0001-8849-5228Le Wu2https://orcid.org/0000-0002-8565-9626Ling Guan3Xun Chen4https://orcid.org/0000-0002-4922-8116Department of Electronic Engineering and Information Science, University of Science and Technology of China, Hefei, ChinaDepartment of Electronic Engineering and Information Science, University of Science and Technology of China, Hefei, ChinaDepartment of Electronic Engineering and Information Science, University of Science and Technology of China, Hefei, ChinaNational Center for Neurological Disorders, Beijing Tiantan Hospital, Beijing, ChinaDepartment of Electronic Engineering and Information Science, University of Science and Technology of China, Hefei, ChinaGeneralized zero-shot learning (GZSL) has significantly reduced the training requirements for steady-state visual evoked potential (SSVEP) based brain-computer interfaces (BCIs). Traditional methods require complete class data sets for training, but GZSL allows for only partial class data sets, dividing them into ‘seen’ (those with training data) and ‘unseen’ classes (those without training data). However, inefficient utilization of SSVEP data limits the accuracy and information transfer rate (ITR) of existing GZSL methods. To this end, we proposed a framework for more effective utilization of SSVEP data at three systematically combined levels: data acquisition, feature extraction, and decision-making. First, prevalent SSVEP-based BCIs overlook the inter-subject variance in visual latency and employ fixed sampling starting time (SST). We introduced a dynamic sampling starting time (DSST) strategy at the data acquisition level. This strategy uses the classification results on the validation set to find the optimal sampling starting time (OSST) for each subject. In addition, we developed a Transformer structure to capture the global information of input data for compensating the small receptive field of existing networks. The global receptive fields of the Transformer can adequately process the information from longer input sequences. For the decision-making level, we designed a classifier selection strategy that can automatically select the optimal classifier for the seen and unseen classes, respectively. We also proposed a training procedure to make the above solutions in conjunction with each other. Our method was validated on three public datasets and outperformed the state-of-the-art (SOTA) methods. Crucially, we also outperformed the representative methods that require training data for all classes.https://ieeexplore.ieee.org/document/10283906/Steady-state visual evoked potentialgeneralized zero-shot learningbrain-computer interface |
| spellingShingle | Xietian Wang Aiping Liu Le Wu Ling Guan Xun Chen Improving Generalized Zero-Shot Learning SSVEP Classification Performance From Data-Efficient Perspective IEEE Transactions on Neural Systems and Rehabilitation Engineering Steady-state visual evoked potential generalized zero-shot learning brain-computer interface |
| title | Improving Generalized Zero-Shot Learning SSVEP Classification Performance From Data-Efficient Perspective |
| title_full | Improving Generalized Zero-Shot Learning SSVEP Classification Performance From Data-Efficient Perspective |
| title_fullStr | Improving Generalized Zero-Shot Learning SSVEP Classification Performance From Data-Efficient Perspective |
| title_full_unstemmed | Improving Generalized Zero-Shot Learning SSVEP Classification Performance From Data-Efficient Perspective |
| title_short | Improving Generalized Zero-Shot Learning SSVEP Classification Performance From Data-Efficient Perspective |
| title_sort | improving generalized zero shot learning ssvep classification performance from data efficient perspective |
| topic | Steady-state visual evoked potential generalized zero-shot learning brain-computer interface |
| url | https://ieeexplore.ieee.org/document/10283906/ |
| work_keys_str_mv | AT xietianwang improvinggeneralizedzeroshotlearningssvepclassificationperformancefromdataefficientperspective AT aipingliu improvinggeneralizedzeroshotlearningssvepclassificationperformancefromdataefficientperspective AT lewu improvinggeneralizedzeroshotlearningssvepclassificationperformancefromdataefficientperspective AT lingguan improvinggeneralizedzeroshotlearningssvepclassificationperformancefromdataefficientperspective AT xunchen improvinggeneralizedzeroshotlearningssvepclassificationperformancefromdataefficientperspective |