Deep Learning on 1-D Biosignals: a Taxonomy-based Survey

Abstract

Our key-code classification of relevant papers was used to cluster the approaches that have been published to date and demonstrated a large variability of research with respect to data, application, and network topology. Future research is expected to focus on the standardization of deep learning architectures and on the optimization of the network parameters to increase performance and robustness. Furthermore, application-driven approaches and updated training data from mobile recordings are needed.

Exploring Deep Physiological Models for Nociceptive Pain Recognition

“…Similar performances have been achieved in the field of 25 speech recognition [8,9] and natural language processing [10,11]. 26 A steadily growing amount of work has been exploring the application of deep 27 learning approaches on physiological signals. Martinéz et al [12] were able to 28 significantly outperform standard approaches built upon hand-crafted features by using 29 a deep learning algorithm for affect modelling based on physiological signals (two 30 physiological signals consisting of Skin Conductance (SC) and Blood Volume Pulse 31 (BVP) were used in this specific work).…”

“…Most of the reported 49 approaches consist of first transforming the processed EMG signal into a two 50 dimensional (time-frequency) visual representation (such as a spectrogram or a 51 scalogram) and subsequently using a deep CNN architecture to proceed with the 52 classification. A similar procedure has been used in [24] for the analysis of deep learning approaches applied to physiological signals can be found in [25] and [26]. 56 The current work focuses on the application of deep learning approaches for 57 nociceptive pain recognition based on physiological signals (EMG, ECG and 58 April 23, 2019 2/16 electrodermal activity (EDA)).…”

Architectural Tricks for Deep Learning in Remote Photoplethysmography

“…Our simple attention model focuses on easy-identifiable parts of the face, known to be important [11,12,14]. CNN expected to use spatial-redundancy as in [23] and extract cross-correlations between signals like in multi-lead ECG [6].…”

“…Therefore features important to filter them out may differs for different heart rates. Also classification models (compared to regression ones, like used by Spetlik et al [28]) are mentioned to be effective for events detection (HR estimation is believed to be based on detection of a heartbeat events) even if the input is noisy [6].…”

“…Facial bounding box is detected by the OpenCV implementation of the Viola-Jones face detector [32] applied to each frame. Regarding the works on ROIs selection [11,12], six ROIs are used in this paper: r = 1..6 ( (1), nose bridge (2), areas under eyes (3,4), truncated facial box (5), full bounding box (6). ROIs coordinates are not ideally symmetrical about a horizontal axis: they were set manually to match the facial features on most video sequences in the dataset.…”