2019/04/11 by Huy P. Phan, Oliver Y. Chén, Phan, Huy +7
Computer Science · Engineering · #FOS: Computer and information sciences #Indoor and Outdoor Localization Technologies #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Music and Audio Processing #Speech and Audio Processing
paper · pdf · doi:10.48550/arxiv.1904.05945
openalex publication_date 2019/04/11 · openalex created_date 2022/07/24 · openalex updated_date 2026/07/28
Many sleep studies suffer from the problem of insufficient data to fully\nutilize deep neural networks as different labs use different recordings set\nups, leading to the need of training automated algorithms on rather small\ndatabases, whereas large annotated databases are around but cannot be directly\nincluded into these studies for data compensation due to channel mismatch. This\nwork presents a deep transfer learning approach to overcome the channel\nmismatch problem and transfer knowledge from a large dataset to a small cohort\nto study automatic sleep staging with single-channel input. We employ the\nstate-of-the-art SeqSleepNet and train the network in the source domain, i.e.\nthe large dataset. Afterwards, the pretrained network is finetuned in the\ntarget domain, i.e. the small cohort, to complete knowledge transfer. We study\ntwo transfer learning scenarios with slight and heavy channel mismatch between\nthe source and target domains. We also investigate whether, and if so, how\nfinetuning entirely or partially the pretrained network would affect the\nperformance of sleep staging on the target domain. Using the Montreal Archive\nof Sleep Studies (MASS) database consisting of 200 subjects as the source\ndomain and the Sleep-EDF Expanded database consisting of 20 subjects as the\ntarget domain in this study, our experimental results show significant\nperformance improvement on sleep staging achieved with the proposed deep\ntransfer learning approach. Furthermore, these results also reveal the\nessential of finetuning the feature-learning parts of the pretrained network to\nbe able to bypass the channel mismatch problem.\n