HASS: High Accuracy Spike Sorting with Wavelet Package Decomposition and Mutual Information
Yao Chen, Libo Huang, Jiong He, Kunyao Zhao, Ruichu Cai, Zhifeng Hao · 2018
Neural signal processing has been dramatically im-proved with the development of microfabrication technology that enables multi-channel signal recording and high-precision signal detection. However, due to the diversity of spike features generated by different neurons and the high noise level in raw data, it is challenging to distinguish the activity of neurons collected by hundreds of closely-positioned recording probes from the background electrical noise. To address this issue, we propose an accurate supervised spike sorting solution based on wavelet package decomposition and mutual information. Furthermore, we build a framework based on the proposed algorithms with a short training stage to automate the data processing. Evaluation results on the raw data from popular datasets show that our solution can provide higher clustering accuracy while maintaining good noise resistance compared to state-of-the-art methods. With our solution, the clustering accuracy can reach up to 99.76% on the dataset with highest noise level. The overall accuracy of our solution can outperform the baseline target by up to 22.35%.