S-Space College of Engineering/Engineering Practice School (공과대학/대학원) Dept. of Mechanical Aerospace Engineering (기계항공공학부) Theses (Ph.D. / Sc.D._기계항공공학부)
Temporalis Muscle Activity Detection through Mechanically Amplified Force Measurement on Glasses
안경에서 기계적으로 증폭된 힘 측정을 통한 측두근 활동의 감지
- 공과대학 기계항공공학부
- Issue Date
- 서울대학교 대학원
- Glasses; law of the lever; wearable device; monitoring of ingestive behavior (MIB); pattern recognition; support vector machine (SVM)
- 학위논문 (박사)-- 서울대학교 대학원 공과대학 기계항공공학부, 2017. 8. 이건우.
- Recently, the form of a pair of glasses is broadly utilized as a wearable device that provides the virtual and augmented reality in addition to its natural functionality as a visual aid. These approaches, however, have lacked the use of its inherent kinematic structure, which is composed of both the temple and the hinge. When we equip the glasses, the force is concentrated at the hinge, which connects the head piece and the temple, from the law of the lever. In addition, since the temple passes through a temporalis muscle, chewing and wink activity, anatomically activated by the contraction and relaxation of the temporalis muscle, can be detected from the mechanically amplified force measurement at the hinge.
This study presents a new and effective method for automatic and objective measurement of the temporalis muscle activity through the natural-born lever mechanism of the glasses. From the implementation of the load cell-integrated wireless circuit module inserted into the both hinges of a 3D printed glasses frame, we developed the system that responds to the temporalis muscle activity persistently regardless of various form factor different from each person. This offers the potential to improve previous studies by avoiding the morphological, behavioral, and environmental constraints of using skin-attached, proximity, and sound sensors. In this study, we collected data featured as sedentary rest, chewing, walking, chewing while walking, talking and wink from 10-subject user study. The collected data were transferred to a series of 84-dimentional feature vectors, each of which was composed of the statistical features of both temporal and spectral domain. These feature vectors, then, were used to define a classifier model implemented by the support vector machine (SVM) algorithm. The model classified the featured activities (chewing, wink, and physical activity) as the average F1 score of 93.7%.
This study provides a novel approach on the monitoring of ingestive behavior (MIB) in a non-intrusive and un-obtrusive manner. It supplies the possibility to apply the MIB into daily life by distinguishing the food intake from the other physical activities such as walking, talking, and wink with higher accuracy and wearability. Furthermore, through applying this approach to a sensor-integrated hair band, it can be potentially used for the medical monitoring of the sleep bruxism or temporomandibular dysfunction.