Audio-based detection and evaluation of eating behavior using the smartwatch platform. (1st October 2015)

• Record Type: Journal Article
• Authors: Kalantarian, Haik; Sarrafzadeh, Majid
• Is Part Of: Computers in biology and medicine. Volume 65(2015)
• Page Start: 1
• Page End: 9
• Publication Date: 2015-10-01
• Subjects: Signal processing -- Wireless Health -- Nutrition -- Smartwatch -- Machine learning -- Pervasive computing; Medicine -- Data processing -- Periodicals; Biology -- Data processing -- Periodicals; 610.285
• DOI: 10.1016/j.compbiomed.2015.07.013
• Languages: English
• ISSNs: 0010-4825

Abstract

Abstract: In recent years, smartwatches have emerged as a viable platform for a variety of medical and health-related applications. In addition to the benefits of a stable hardware platform, these devices have a significant advantage over other wrist-worn devices, in that user acceptance of watches is higher than other custom hardware solutions. In this paper, we describe signal-processing techniques for identification of chews and swallows using a smartwatch device׳s built-in microphone. Moreover, we conduct a survey to evaluate the potential of the smartwatch as a platform for monitoring nutrition. The focus of this paper is to analyze the overall applicability of a smartwatch-based system for food-intake monitoring. Evaluation results confirm the efficacy of our technique; classification was performed between apple and potato chip bites, water swallows, talking, and ambient noise, with an F -measure of 94.5% based on 250 collected samples. Abstract : Highlights: We evaluate the potential of a smartwatch for monitoring eating. We use the microphone of the Samsung Galaxy Gear to record audio samples. We use the RandomForest classifier to classify between several foods. We are able to discriminate between eating and talking.