@inproceedings{oai:ipsj.ixsq.nii.ac.jp:00241864,
 author = {Soma, Kato and Midori, Sugaya and Soma, Kato and Midori, Sugaya},
 book = {Proceedings of Asia Pacific Conference on Robot IoT System Development and Platform},
 month = {Dec},
 note = {In recent years, with the widespread adoption of IoT, the number of devices connected to networks has increased, leading to the generation of vast amounts of sensing data. High-performance and high-efficiency processing of this data is increasingly expected. One application that handles such large volumes of sensing data is EEG (electroencephalography). EEG is widely utilized not only for medical purposes but also in research fields such as cognitive science, psychology, and physiology, as well as in areas like neuromarketing. Its use is anticipated to expand further in the future. However, there are several challenges when it comes to efficiently processing EEG data. The first challenge is that while individual algorithms have been accelerated, general-purpose acceleration methods have not been sufficiently discussed. The second challenge is that there is no adequately discussed, straightforward method for speeding up the calculation of EEG features, leading to time-consuming processes. This study aims to accelerate EEG processing and reduce the effort required by parallelizing the calculation of EEG features. Additionally, as a method that allows for easy description, we propose an automatic parallelization method that utilizes YAML for describing the processing flow. Multiple parallelization methods were implemented, and their performance was evaluated by comparing processing times. As a result, the Buffering method, in particular, achieved up to approximately 9 times faster processing compared to sequential processing., In recent years, with the widespread adoption of IoT, the number of devices connected to networks has increased, leading to the generation of vast amounts of sensing data. High-performance and high-efficiency processing of this data is increasingly expected. One application that handles such large volumes of sensing data is EEG (electroencephalography). EEG is widely utilized not only for medical purposes but also in research fields such as cognitive science, psychology, and physiology, as well as in areas like neuromarketing. Its use is anticipated to expand further in the future. However, there are several challenges when it comes to efficiently processing EEG data. The first challenge is that while individual algorithms have been accelerated, general-purpose acceleration methods have not been sufficiently discussed. The second challenge is that there is no adequately discussed, straightforward method for speeding up the calculation of EEG features, leading to time-consuming processes. This study aims to accelerate EEG processing and reduce the effort required by parallelizing the calculation of EEG features. Additionally, as a method that allows for easy description, we propose an automatic parallelization method that utilizes YAML for describing the processing flow. Multiple parallelization methods were implemented, and their performance was evaluated by comparing processing times. As a result, the Buffering method, in particular, achieved up to approximately 9 times faster processing compared to sequential processing.},
 pages = {22--28},
 publisher = {情報処理学会},
 title = {Implementation and Evaluation of Python's Automatic Parallelization Function for Accelerating EEG Feature Extraction},
 volume = {2024},
 year = {2024}
}