In recent years, many people have owned wearable devices such as smartphones and smartwatches, in which the built-in sensors were installed. Moreover, such kind of build sensors has been used for various researches like height estimation, road condition estimation, and so on. In addition, it also conducted a behavioral estimation of smartphone-owners, possession location estimation, and person estimation. However, in the case of behavioral estimation, most of them use data measured by fixing the possession position at a single location, and behavioral estimation that takes the possession position of the terminal into account has not been conducted. It is thought that users possess terminals such as smartphones in multiple locations. Therefore, the purpose of this study is to perform a behavioral estimation that considers the multiple possession locations of a user's smartphone. Also, suppose it is possible to estimate a user's behavior by considering the multiple possession locations of the smartphones. In that case, it may be possible to change notifications such as emails and phone calls on smartphones according to the user's state. As a result of classifying 12 classes using LSTM[1], a combination of the GAF[2] algorithm and ResNet50[3], and a combination of the GAF algorithm and Mdk-ResNet[4], the combination of the GAF algorithm and Mdk-ResNet had the highest accuracy, with Accuracy: 98 .892%.
Although lip-reading using image processing and machine learning have been performed at the word level, LipNet, (1) a network that enables recognition at the sentence level, improves the recognition accuracy over the former method. However, this was the case for English speakers, and no extra experimental result has been reported for Japanese speakers. This study aims to create an experimental database for Japanese speech scenes containing all 50 Japanese sounds and evaluate the recognition accuracy using LipNet. (1)
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.