TY - GEN
T1 - Motorcyclists' head motions recognition by using the smart helmet with low sampling rate
AU - Chen, Yu Ren
AU - Tsai, Chang Ming
AU - Wong, Ka Io
AU - Lee, Tzu Chang
AU - Loh, Chee Hoe
AU - Ying, Jia Ching
AU - Chen, Yi Chung
N1 - Publisher Copyright:
© 2019 IEEE.
PY - 2019/8
Y1 - 2019/8
N2 - The number of traffic incidents involving motorcyclists is on the rise; consequently research has focused on analysis of the head motions of motorcyclists to determine their level of concentration on the road while driving. These studies used three-axis accelerometers in helmets to record the acceleration signals that are detected when motorcyclists move their heads and then analyzed these signals using machine learning. However, we found that these methods are not very effective for the following reasons: (1) battery and memory capacity constraints mean that helmet sensors cannot collect acceleration data frequently, so the results cannot completely present head motions. (2) When motorcyclists are riding, the acceleration data collected by the helmets not only include the acceleration data of motorcyclist head motions but also include the acceleration data of motorcycle movement, which creates difficulties for recognition. (3) Due to the volume constraints of helmets, we cannot install GPUs or large-capacity batteries, so more complex models or deep learning models cannot be directly used for head motion recognition. (4) Head motions are smaller than body or limb motions, and most head motions do not occur periodically, which makes recognition even more difficult. To overcome these issues, this study proposed a novel machine learning method combined with a fuzzy neural network to perform motorcyclist head motion recognition with low-frequency acceleration signals collected from helmets. Experiment simulations demonstrate the validity of the proposed method.
AB - The number of traffic incidents involving motorcyclists is on the rise; consequently research has focused on analysis of the head motions of motorcyclists to determine their level of concentration on the road while driving. These studies used three-axis accelerometers in helmets to record the acceleration signals that are detected when motorcyclists move their heads and then analyzed these signals using machine learning. However, we found that these methods are not very effective for the following reasons: (1) battery and memory capacity constraints mean that helmet sensors cannot collect acceleration data frequently, so the results cannot completely present head motions. (2) When motorcyclists are riding, the acceleration data collected by the helmets not only include the acceleration data of motorcyclist head motions but also include the acceleration data of motorcycle movement, which creates difficulties for recognition. (3) Due to the volume constraints of helmets, we cannot install GPUs or large-capacity batteries, so more complex models or deep learning models cannot be directly used for head motion recognition. (4) Head motions are smaller than body or limb motions, and most head motions do not occur periodically, which makes recognition even more difficult. To overcome these issues, this study proposed a novel machine learning method combined with a fuzzy neural network to perform motorcyclist head motion recognition with low-frequency acceleration signals collected from helmets. Experiment simulations demonstrate the validity of the proposed method.
UR - http://www.scopus.com/inward/record.url?scp=85083438932&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85083438932&partnerID=8YFLogxK
U2 - 10.1109/Ubi-Media.2019.00038
DO - 10.1109/Ubi-Media.2019.00038
M3 - Conference contribution
AN - SCOPUS:85083438932
T3 - Proceedings - 2019 12th International Conference on Ubi-Media Computing, Ubi-Media 2019
SP - 157
EP - 163
BT - Proceedings - 2019 12th International Conference on Ubi-Media Computing, Ubi-Media 2019
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 12th International Conference on Ubi-Media Computing, Ubi-Media 2019
Y2 - 6 August 2019 through 9 August 2019
ER -