Watch-Dog: Detecting Self-Harming Activities From Wrist Worn Accelerometers

Abstract

In a 2012 survey, in the United States alone, there were more than 35 000 reported suicides with approximately 1800 of being psychiatric inpatients. Recent Centers for Disease Control and Prevention (CDC) reports indicate an upward trend in these numbers. In psychiatric facilities, staff perform intermittent or continuous observation of patients manually in order to prevent such tragedies, but studies show that they are insufficient, and also consume staff time and resources. In this paper, we present the Watch-Dog system, to address the problem of detecting self-harming activities when attempted by in-patients in clinical settings. Watch-Dog comprises of three key components-Data sensed by tiny accelerometer sensors worn on wrists of subjects; an efficient algorithm to classify whether a user is active versus dormant (i.e., performing a physical activity versus not performing any activity); and a novel decision selection algorithm based on random forests and continuity indices for fine grained activity classification. With data acquired from 11 subjects performing a series of activities (both self-harming and otherwise), Watch-Dog achieves a classification accuracy of , , and for same-user 10-fold cross-validation, cross-user 10-fold cross-validation, and cross-user leave-one-out evaluation, respectively. We believe that the problem addressed in this paper is practical, important, and timely. We also believe that our proposed system is practically deployable, and related discussions are provided in this paper.

A Portable Sign Language Collection and Translation Platform with Smart Watches Using a BLSTM-Based Multi-Feature Framework

“…During the experiment, one signer was randomly selected as the testing signer, while other signers were considered as training data. For this paper, five types of machine learning approaches [27,32,33,36,46] were selected for comparison, including the support vector machine (SVM), random forest (RF), K-nearest neighbors (KNN), linear discriminant analysis (LDA), and Gaussian mixture model (GMM).…”

Systematic review of interventions to improve constant observation on adult inpatient psychiatric wards

“…Studies were categorized based on the main intervention components (see Box 1) and whether the intervention made changes directly to constant observation or made systemic organizational changes on the ward. Out of the eight descriptive interventions, three could be categorized into a single component (Bharti et al ., ; Björkdahl et al ., ; Janofsky, ) and the remaining interventions were complex consisting of two or more components (Ashaye, Ikkos & Rigby, ; Dennis, ; Pereira & Woollaston, ; Russ, ; Sullivan & Rivera, ). The most commonly used intervention components were staff education and training, and changes to record keeping and assessment, and one intervention described the use of technology as an alternative tool to constant observation.…”

CovidAlert -- A Wristwatch-based System to Alert Users from Face Touching

“…Currently, Apple and Android watch users can precisely track various activities, including stand, walk, run and other physical exercises throughout the day. In research, smartwatch-based HAR system have been explored for elderly assistance [11], detection of self-harming activities in psychiatric facilities [12], distracted driving [13], smoking activities [14], speed detection [15] and more. In this section, we have discussed only recent works related to face touching activities in the context of Covid-19 pandemic.…”

“…STA/LTA algorithm is used in seismology to detect the intensity of earthquakes [19]. In healthcare application, Bharti et al [12] have leveraged the algorithm to detect self-harm activity in psychiatric facilities.…”