{"id":{"repo_id":"uts","oai_identifier":"oai:opus.lib.uts.edu.au:10453/164468"},"canonical_url":"https://search.dev.ndltd.org/etd/uts/oai:opus.lib.uts.edu.au:10453/164468","repository":{"repo_id":"uts","name":"University of Technology Sydney","base_url":"https://opus.lib.uts.edu.au/oai/request"},"display":{"title":"Spiking Neural Network Framework for Brain Computer Interfaces","abstract":"In the field of Brain-computer interfaces (BCI), the majority of studies often rely on electroencephalography (EEG) over invasive methods despite the influence of noise and insufficient spatial resolution due to EEG's simplicity for practical usage. However, with EEG being an accumulation of spiking neural activity, reverse engineering it to a spiking version could establish new directions in improving BCI systems, which is the focus of this thesis. Spiking neural network (SNN) communicates via spikes and is a suitable model to study the spiking aspect of EEG-based BCI systems. Electroencephalography recordings are often time consuming, and artifacts cause further rejection of these recorded samples. To overcome this, a few studies have developed generative models based on deep neural network (DNN) to create artificial EEG samples, but these models still require a huge number of samples which is counter-intuitive. Therefore, in this study, an SNN-based method was developed that can reconstruct EEG and generate synthetic samples with few original samples. These synthetic samples improved the motor imagery (MI) performance. To further explore the utility of SNN, it was trained to learn the template of an EEG signal and in turn obtain an approximated spike representation of the EEG samples. This significantly enhanced the classification performance in comparison to the baseline EEG-based performance of P300 and error-related negativity (ERN). The efficacy of the SNN depended on the neuron properties to an extent, and leveraging the heterogeneity of the properties could provide a better spike representation of the EEG. For ease of computation, the majority of past studies relied on a simplified spiking neuron model that could only produce tonic spiking. Therefore, we introduced a new spiking model termed the burst spiking neural unit (BSNU) that could produce both tonic spiking and bursting signals. The BSNU model was initially validated on different machine learning benchmark tests to assess the efficacy of bursting and then the model was explored towards its ability for spike-representation. The BSNU-based SNN model outperformed SNU on the benchmark tests and has also been shown to further improve spike representation of EEG signals. This resulted in further enhancing the classification performance of P300 and MI datasets. In summary, this thesis is a consolidation of the SNN's role in improving different aspects of BCI paradigms, from data augmentation, and feature extraction to classification performance.","abstract_html":"In the field of Brain-computer interfaces (BCI), the majority of studies often rely on electroencephalography (EEG) over invasive methods despite the influence of noise and insufficient spatial resolution due to EEG's simplicity for practical usage. However, with EEG being an accumulation of spiking neural activity, reverse engineering it to a spiking version could establish new directions in improving BCI systems, which is the focus of this thesis. Spiking neural network (SNN) communicates via spikes and is a suitable model to study the spiking aspect of EEG-based BCI systems. Electroencephalography recordings are often time consuming, and artifacts cause further rejection of these recorded samples. To overcome this, a few studies have developed generative models based on deep neural network (DNN) to create artificial EEG samples, but these models still require a huge number of samples which is counter-intuitive. Therefore, in this study, an SNN-based method was developed that can reconstruct EEG and generate synthetic samples with few original samples. These synthetic samples improved the motor imagery (MI) performance. To further explore the utility of SNN, it was trained to learn the template of an EEG signal and in turn obtain an approximated spike representation of the EEG samples. This significantly enhanced the classification performance in comparison to the baseline EEG-based performance of P300 and error-related negativity (ERN). The efficacy of the SNN depended on the neuron properties to an extent, and leveraging the heterogeneity of the properties could provide a better spike representation of the EEG. For ease of computation, the majority of past studies relied on a simplified spiking neuron model that could only produce tonic spiking. Therefore, we introduced a new spiking model termed the burst spiking neural unit (BSNU) that could produce both tonic spiking and bursting signals. The BSNU model was initially validated on different machine learning benchmark tests to assess the efficacy of bursting and then the model was explored towards its ability for spike-representation. The BSNU-based SNN model outperformed SNU on the benchmark tests and has also been shown to further improve spike representation of EEG signals. This resulted in further enhancing the classification performance of P300 and MI datasets. In summary, this thesis is a consolidation of the SNN's role in improving different aspects of BCI paradigms, from data augmentation, and feature extraction to classification performance.","abstract_has_math":false,"creators":["Singanamalla, Sai Kalyan Ranga"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2022,"date_issued":"2022","date_published":"2022","updated_at":"2026-07-24T06:32:42Z","subjects":[],"languages":["en_US"],"rights":["The author owns the copyright in this thesis including all reproduction and reuse rights for the work. The work may not be altered without the permission of the copyright owner. Attribution is essential when quoting or paraphrasing from this thesis.","au.edu.uts.lib/ppc","info:eu-repo/semantics/openAccess"],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"http://hdl.handle.net/10453/164468","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:creator","label":"Author","values":["Singanamalla, Sai Kalyan Ranga"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2022-12-18T23:29:21Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2022-12-18T23:29:21Z"]},{"key":"dc:date.issued","label":"Date","values":["2022"]},{"key":"dc:relation","label":"Dc Relation","values":["https://opus.lib.uts.edu.au/bitstream/10453/164468/2/02whole.pdf"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en_US"]},{"key":"dc:rights","label":"Dc Rights","values":["The author owns the copyright in this thesis including all reproduction and reuse rights for the work. The work may not be altered without the permission of the copyright owner. Attribution is essential when quoting or paraphrasing from this thesis.","au.edu.uts.lib/ppc","info:eu-repo/semantics/openAccess"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["http://hdl.handle.net/10453/164468"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["University of Technology Sydney. Faculty of Engineering and Information Technology."]},{"key":"dc:description.abstract","label":"Abstract","values":["In the field of Brain-computer interfaces (BCI), the majority of studies often rely on electroencephalography (EEG) over invasive methods despite the influence of noise and insufficient spatial resolution due to EEG's simplicity for practical usage. However, with EEG being an accumulation of spiking neural activity, reverse engineering it to a spiking version could establish new directions in improving BCI systems, which is the focus of this thesis. Spiking neural network (SNN) communicates via spikes and is a suitable model to study the spiking aspect of EEG-based BCI systems. Electroencephalography recordings are often time consuming, and artifacts cause further rejection of these recorded samples. To overcome this, a few studies have developed generative models based on deep neural network (DNN) to create artificial EEG samples, but these models still require a huge number of samples which is counter-intuitive. Therefore, in this study, an SNN-based method was developed that can reconstruct EEG and generate synthetic samples with few original samples. These synthetic samples improved the motor imagery (MI) performance. To further explore the utility of SNN, it was trained to learn the template of an EEG signal and in turn obtain an approximated spike representation of the EEG samples. This significantly enhanced the classification performance in comparison to the baseline EEG-based performance of P300 and error-related negativity (ERN). The efficacy of the SNN depended on the neuron properties to an extent, and leveraging the heterogeneity of the properties could provide a better spike representation of the EEG. For ease of computation, the majority of past studies relied on a simplified spiking neuron model that could only produce tonic spiking. Therefore, we introduced a new spiking model termed the burst spiking neural unit (BSNU) that could produce both tonic spiking and bursting signals. The BSNU model was initially validated on different machine learning benchmark tests to assess the efficacy of bursting and then the model was explored towards its ability for spike-representation. The BSNU-based SNN model outperformed SNU on the benchmark tests and has also been shown to further improve spike representation of EEG signals. This resulted in further enhancing the classification performance of P300 and MI datasets. In summary, this thesis is a consolidation of the SNN's role in improving different aspects of BCI paradigms, from data augmentation, and feature extraction to classification performance."]},{"key":"dc:format","label":"Dc Format","values":["Thesis (PhD)"]},{"key":"dc:title","label":"Title","values":["Spiking Neural Network Framework for Brain Computer Interfaces"]}]}],"canonical_facts":{"dc:creator":["Singanamalla, Sai Kalyan Ranga"],"dc:date.accessioned":["2022-12-18T23:29:21Z"],"dc:date.available":["2022-12-18T23:29:21Z"],"dc:date.issued":["2022"],"dc:description":["University of Technology Sydney. Faculty of Engineering and Information Technology."],"dc:description.abstract":["In the field of Brain-computer interfaces (BCI), the majority of studies often rely on electroencephalography (EEG) over invasive methods despite the influence of noise and insufficient spatial resolution due to EEG's simplicity for practical usage. However, with EEG being an accumulation of spiking neural activity, reverse engineering it to a spiking version could establish new directions in improving BCI systems, which is the focus of this thesis. Spiking neural network (SNN) communicates via spikes and is a suitable model to study the spiking aspect of EEG-based BCI systems. Electroencephalography recordings are often time consuming, and artifacts cause further rejection of these recorded samples. To overcome this, a few studies have developed generative models based on deep neural network (DNN) to create artificial EEG samples, but these models still require a huge number of samples which is counter-intuitive. Therefore, in this study, an SNN-based method was developed that can reconstruct EEG and generate synthetic samples with few original samples. These synthetic samples improved the motor imagery (MI) performance. To further explore the utility of SNN, it was trained to learn the template of an EEG signal and in turn obtain an approximated spike representation of the EEG samples. This significantly enhanced the classification performance in comparison to the baseline EEG-based performance of P300 and error-related negativity (ERN). The efficacy of the SNN depended on the neuron properties to an extent, and leveraging the heterogeneity of the properties could provide a better spike representation of the EEG. For ease of computation, the majority of past studies relied on a simplified spiking neuron model that could only produce tonic spiking. Therefore, we introduced a new spiking model termed the burst spiking neural unit (BSNU) that could produce both tonic spiking and bursting signals. The BSNU model was initially validated on different machine learning benchmark tests to assess the efficacy of bursting and then the model was explored towards its ability for spike-representation. The BSNU-based SNN model outperformed SNU on the benchmark tests and has also been shown to further improve spike representation of EEG signals. This resulted in further enhancing the classification performance of P300 and MI datasets. In summary, this thesis is a consolidation of the SNN's role in improving different aspects of BCI paradigms, from data augmentation, and feature extraction to classification performance."],"dc:format":["Thesis (PhD)"],"dc:identifier.uri":["http://hdl.handle.net/10453/164468"],"dc:language.iso":["en_US"],"dc:relation":["https://opus.lib.uts.edu.au/bitstream/10453/164468/2/02whole.pdf"],"dc:rights":["The author owns the copyright in this thesis including all reproduction and reuse rights for the work. The work may not be altered without the permission of the copyright owner. Attribution is essential when quoting or paraphrasing from this thesis.","au.edu.uts.lib/ppc","info:eu-repo/semantics/openAccess"],"dc:title":["Spiking Neural Network Framework for Brain Computer Interfaces"],"dc:type":["Thesis"]},"updated_at":"2026-07-24T06:32:42Z"}