山口さんはこの発表で「IWIN2023 Student Awards」を受賞しました
熊澤「Analysis and Sharing of Cooking Actions Using Wearable Sensors」
There are many different ways to share cooking recipes and pervasive in the world. Now that the Internet has become widespread and information sharing has become easier, recipes can be browsing on the Internet. Mechanisms already exist for sharing dish quality and taste through photos and text. On the other hand, there is no mechanism to compare cooking ability and efficiency with past selves and others. This research aims to increase the enjoyment of cooking by sharing and comparing recipes and cooking actions. The approach is to use wearable sensors to sense the cooking process and extract features for comparison. The contribution of this research is twofold. The first is a proposal for a framework for a system that allows compare cooking ability and efficiency with past selves and others. The second is the actual sensing of cooking behavior and providing concrete examples of analysis and comparison with respect to transitions in behavior and place.
この度 9/1 ~ 9/4 に開催されたIWIN2023にオフラインで参加してきました．
山口「Method for Estimating Bicycle Air Pressure Decrease based on Vibration Sensing of Bicycles using Smartphone」
Because of the lack of bicycle inspection before driving, some people experience accidents or near misses during driving. There is a need to prompt drivers to conduct bicycle inspection to reduce the number of people who experience this. In this study, we focus on air pressure inspection and propose a method for estimating bicycle air pressure decrease based on vibration sensing using smartphone. The acceleration in the direction perpendicular to the ground is acquired with asmartphone, and the standard deviation and the amplitude spectrum of each frequency are used as features to estimate the air pressure decrease using random forests and other methods. Experiments were conducted to classify whether or not air inflation is necessary, to estimate the air pressure value and classify it based on that value, and to evaluate whether the training data from other locations and bicycles can be diverted. The results of the evaluation experiments showed that 96.1% correctly classified when inflation was required, but it was difficult to reuse the training data.