Daido University

About: Daido University is a education organization based out in Nagoya, Japan. It is known for research contribution in the topics: Ultimate tensile strength & Proton exchange membrane fuel cell. The organization has 209 authors who have published 423 publications receiving 3223 citations.

Effects of Hydrogen on Frictional Properties of DLC Films

Abstract: The tribological behavior of diamond-like carbon films (DLC) is strongly dependent on the hydrogen content, sp/sp ratio, and sliding environment. Some hydrogenated amorphous carbon films (a-C:H) exhibit superlow friction in hydrogen conditions. However, previous works have not clarified the dominant factors of the superlow friction phenomena of DLC films. In this research, we focused on the effects of hydrogen derived from the surrounding atmosphere and the hydrogen within the DLC films on superlow friction phenomena. To investigate these effects, friction tests were conducted on three DLC films having different hydrogen contents (0, and 18, 30 at%) in the air and in low-pressure-hydrogen conditions at various hydrogen pressures. After the friction tests, the wear tracks were examined by confocal laser scanning microscopy, Raman spectroscopy, elastic recoil detection (ERDA) analysis, and time-of-flight secondary ion mass spectrometry (TOF-SIMS). The hydrogen derived from the surrounding atmosphere and the formation of the hydrogen-rich tribofilm were key factors for the superlow friction phenomena.

Risky Action Recognition in Lane Change Video Clips using Deep Spatiotemporal Networks with Segmentation Mask Transfer

Abstract: Advanced driver assistance and automated driving systems rely on risk estimation modules to predict and avoid dangerous situations. Current methods use expensive sensor setups and complex processing pipeline, limiting their availability and robustness. To address these issues, we introduce a novel deep learning based action recognition framework for classifying dangerous lane change behavior in short video clips captured by a monocular camera. We designed a deep spatiotemporal classification network that uses pre-trained state-of-the-art instance segmentation network Mask R-CNN as its spatial feature extractor for this task. The Long-Short Term Memory (LSTM) and shallower final classification layers of the proposed method were trained on a semi-naturalistic lane change dataset with annotated risk labels. A comprehensive comparison of state-of-the-art feature extractors was carried out to find the best network layout and training strategy. The best result, with a 0.937 AUC score, was obtained with the proposed network. Our code and trained models are available open-source.

Personal identification method using footsteps

Abstract: In recent years, personal authentications using biological information are used for protection of personal data and confidential information in local governments and companies. In this paper, we propose a novel personal identification method using footsteps. The users' mental burdens of the proposal technique is a little because the footsteps can be easily recorded without special equipments. First, the proposed method detects footstep sections from the recorded signals. Then, the acoustic feature parameters, which are Mel-Frequency Cepstral Coefficients (MFCCs), ΔMFCCs, and ΔLogarithm Powers (ΔLPs), are extracted as footstep features from the footstep section. Finally, persons are identified by k-Nearest Neighbor (k-NN) in which Dynamic Programming matching algorithm (DP) is used as a distance measure and/or Gaussian Mixture Models (GMMs). We conduct personal identification experiments using 720 footstep data which are recorded from 12 test subjects for evaluating the proposed method. From the experimental results, average accuracies of overall footwear are 79.9% and 92.8% in k-NN and GMMs.