ソンラートラムワニッチ ウィラット

The Thammasat AI City distributed platform is a proposed AI platform designed to enhance city intelligent management. It addresses the limitations of current smart city architecture by incorporating cross-domain data connectivity and machine learning to support comprehensive data collection. In this study, we delve into two main areas, that is, monitoring and visualization of city ambient lighting, and indoor human physical distance tracking. The smart street light monitoring system provides real-time visualization of street lighting status, energy consumption, and maintenance requirement, which helps to optimize energy consumption and maintenance reduction. The indoor camera-based system for human physical distance tracking can be used in public spaces to monitor social distancing and ensure public safety. The overall goal of the platform is to improve the quality of life in urban areas and align with sustainable urban development concepts.

The smart city concept has been popularized in the urbanization of major metropolitan areas through the implementation of intelligent systems and technology to serve the increasing human population. This work developed an automatic light adjustment system at Thammasat University, Rangsit Campus, Thailand, with a primary objective of optimizing energy efficiency, while providing sufficient illumination for the campus. The development consists of two sections: the device control and the prediction model. The device control functionalities were developed with the user interface to enable control of the smart street light devices and the application programming interface (API) to send the light-adjusting command. The prediction model was created using an AI-assisted data analytic platform to obtain the predicted illuminance values so as to, subsequently, suggest light-dimming values according to the current environment. Four machine-learning models were performed on a nine-month environmental dataset to acquire predictions. The result demonstrated that the three-day window size setting with the XGBoost model yielded the best performance, attaining the correlation coefficient value of 0.922, showing a linear relationship between actual and predicted illuminance values using the test dataset. The prediction retrieval API was established and connected to the device control API, which later created an automated system that operated at a 20-min interval. This allowed real-time feedback to automatically adjust the smart street lighting devices through the purpose-designed data analytics features.

This paper describes about project “Data Sensorium” launched at the Asia AI Institute of Musashino University. Data Sensoriumis a conceptual framework of systems providing physical experience of content stored in database. Spatial immersive display is a key technology of Data Sensorium. This paper introduces prototype implementation of the concept and its application to environmental and architectural dataset.

Through technology, it is essential to seamlessly bridge the divide between diverse speaking communities (including the signer (the sign language speaker) community). In order to realize communication that successfully conveys emotions, it is necessary to recognize not only verbal information but also non-verbal information. In the case of signers, there are two main types of behavior: verbal behavior and emotional behavior. This paper presents a sign language recognition method by similarity measure with emotional expression specific to signers. We focus on recognizing the sign language conveying verbal information itself and on recognizing emotional expression. Our method recognizes sign language by time-series similarity measure on a small amount of model data, and at the same time, recognizes emotion expression specific to signers. Our method extracts time-series features of the body, arms, and hands from sign language videos and recognizes them by measuring the similarity of the time-series features. In addition, it recognizes the emotional expressions specific to signers from the time-series features of their faces.

This research proposes an AI platform for data sharing across multiple domains. Since the data in the smart city concept are domain-specific processed, the existing smart city architecture is suffered from cross-domain data interpretation. To go beyond the digital transformation efforts in smart city development, the AI city is created on the architecture of cross-domain data connectivity and transform learning in the machine learning paradigm. In this research, the health and human behavioral data are targeted on human traceability and contactless technologies. To measure the inhabitants quality of life (QoL), the primary emotion expression study is conducted to interpret the emotional states and the mental health of people in the urbanized city. The results of information augmentation draw attention to the immersive visualization of the Thammasat model.

© 2019 IEEE. This paper describes an approach for bed position classification by using 2 stacked layers of Long Short-Term Memory approach. The data is collected from the sensor panel which consists of 2 types of sensors, i.e. piezoelectric and pressure sensors. The raw data has been classified into 5 classes. It also has to go through the min-max scaling normalization on a fixed range between 0 and 1. The data is assembled to fit a one-second interval of the 30Hz sensor sampling rate. The model has been experimented by changing the number of hidden nodes of the model in 128, 80 and 50 nodes. The result is 91.70% of accuracy which is good enough comparing to the previous works.

© 2018 IEEE. In the Thai language, named entity can be used with or without a prefix or an indication of word. This may cause confusion between named entity and other types of noun. However, a named entity is likely to be used in adjacent to verbs or prepositions. This means that the adjacent verbs or prepositions to a noun can be as a good feature to determine the type of named entity. There are some studies on named entity recognition (NER) task in other languages such as Indonesian showing that combination of word embedding and part-of-speech (POS) tag can improve the performance of the NER model. In this paper, we investigate the Thai Named Entity Recognition task using Bi-LSTM model with word embedding and POS embedding for dealing with the relatively small and disjointedly labeled corpus. We compare our model with the one without POS tag, and the baseline model of CRF with the similar set of feature. The experiment results show that our proposed model outperforms the other two in all F1-score measures. Especially, in the case of location file, the F1-score is increased by 14 percent.

Copyright © 2018 The Institute of Electronics, Information and Communication Engineers. This paper presents a supervised method to classify a document at the sub-sentence level. Traditionally, sentiment analysis often classifies sentence polarity based on word features, syllable features, or Ngram features. A sentence, as a whole, may contain several phrases and words which carry their own specific sentiment. However, classifying a sentence based on phrases and words can sometimes be incoherent because they are ungrammatically formed. In order to overcome this problem, we need to arrange words and phrase in a dependency form to capture their semantic scope of sentiment. Thus, we transform a sentence into a dependency tree structure. A dependency tree is composed of subtrees, and each subtree allocates words and syllables in a grammatical order. Moreover, a sentence dependency tree structure can mitigate word sense ambiguity or solve the inherent polysemy of words by determining their word sense. In our experiment, we provide the details of the proposed subtree polarity classification for sub-opinion analysis. To conclude our discussion, we also elaborate on the effectiveness of the analysis result.

© 2018 The authors and IOS Press. All rights reserved. This study proposes bed posture classification using a Neural Network and a Bayesian Network for elderly care. The data are collected in a hospital. The on-bed postures are analyzed into five types, those are, out of bed, sitting, lying down, lying left, and lying right, by using signals from a sensor panel (composed of piezoelectric sensors and pressure sensors). The sensor panel is placed under a mattress in the thoracic area. To eliminate the effect of weight and the bias between different types of sensors, the sensing data are normalized into a range of 0 to 1 by the unity-based normalization (or feature scaling) method. In addition, a Bayesian Network is adopted to estimate the likelihood of consecutive postures. The results from both a Neural Network and Bayesian Network estimation are combined by the weighted arithmetic mean. The experimental results yield the maximum accuracy of posture classification when the coefficient of Bayesian probability and a Neural Network are set to 0.7 and 0.3 respectively.

© 2018 The authors and IOS Press. All rights reserved. In the last few years, Concept Similarity Measures (CSMs) become important for the biomedical ontologies in order to find adaptable treatments from the conceptually similar diseases. For the ontology primitive concepts, they are not fully defined in the ontology so taxonomical path-based similarity measure cannot give the correct similarity for primitive concepts. In this paper, we propose a new primitive concept name similarity measure based on natural language processing to get a better result in concept similarity measure in terms of noun phrase construction analysis. We conduct experiments on the standard clinical ontology SNOMED CT and make comparison between taxonomical path-based measure and our proposed similarity measure against human expert results in order to prove our proposed similarity measure can outperform the existing approaches for primitive concept similarity.