We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

eess.SP

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Electrical Engineering and Systems Science > Signal Processing

Title: Binary Single-dimensional Convolutional Neural Network for Seizure Prediction

Abstract: Nowadays, several deep learning methods are proposed to tackle the challenge of epileptic seizure prediction. However, these methods still cannot be implemented as part of implantable or efficient wearable devices due to their large hardware and corresponding high-power consumption. They usually require complex feature extraction process, large memory for storing high precision parameters and complex arithmetic computation, which greatly increases required hardware resources. Moreover, available yield poor prediction performance, because they adopt network architecture directly from image recognition applications fails to accurately consider the characteristics of EEG signals. We propose in this paper a hardware-friendly network called Binary Single-dimensional Convolutional Neural Network (BSDCNN) intended for epileptic seizure prediction. BSDCNN utilizes 1D convolutional kernels to improve prediction performance. All parameters are binarized to reduce the required computation and storage, except the first layer. Overall area under curve, sensitivity, and false prediction rate reaches 0.915, 89.26%, 0.117/h and 0.970, 94.69%, 0.095/h on American Epilepsy Society Seizure Prediction Challenge (AES) dataset and the CHB-MIT one respectively. The proposed architecture outperforms recent works while offering 7.2 and 25.5 times reductions on the size of parameter and computation, respectively.
Comments: 2020 IEEE International Symposium on Circuits and Systems (ISCAS)
Subjects: Signal Processing (eess.SP); Human-Computer Interaction (cs.HC); Machine Learning (cs.LG)
DOI: 10.1109/ISCAS45731.2020.9180430
Cite as: arXiv:2206.07518 [eess.SP]
  (or arXiv:2206.07518v1 [eess.SP] for this version)

Submission history

From: Shiqi Zhao [view email]
[v1] Wed, 8 Jun 2022 09:27:37 GMT (1659kb,D)

Link back to: arXiv, form interface, contact.