Home // SIGNAL 2022, The Seventh International Conference on Advances in Signal, Image and Video Processing // View article

Alcohol Detection over Long Periods Using Smartphone Accelerometer Data

Authors:
Manuel Gil-Martín
Rubén San-Segundo
Cristina Luna-Jiménez

Keywords: Alcohol Detection; Motion Wearable Sensors; Convolutional Neural Networks; Sub-windows Combination

Abstract:
This paper proposes a motion biomarker for alcohol detection using a deep learning approach that processes inertial signals recorded with a smartphone. The deep learning architecture is composed of a Convolutional Neural Network, including three convolutional layers for learning features from the inertial signal spectrum, and several fully connected layers to perform classification and regression tasks. The motion biomarker is computed in two steps. Firstly, the inertial signals are segmented in short sub-windows (3-6 seconds) and the system generates a score for each sub-window. Secondly, the scores in consecutive sub-windows are combined to provide a motion biomarker over in longer periods of time (30 seconds). This paper compares the proposed approach to previous works using the same experimental dataset and setup: Bar Crawl Detecting Heavy Drinking Data Set, K-fold cross-validation methodology and two tasks (classification and regression). The proposed deep learning approach overperformed previous reported results: the accuracy increased 4 % (absolute) when classifying between intoxicated and sober participants and the Mean Squared Error relatively decreased 9 % when estimating the Transdermal Alcohol Content of the participants by averaging the scores from consecutive sub-windows.

Pages: 13 to 17

Copyright: Copyright (c) IARIA, 2022

Publication date: May 22, 2022

Published in: conference

ISSN: 2519-8432

ISBN: 978-1-61208-970-6

Location: Venice, Italy

Dates: from May 22, 2022 to May 26, 2022