Masanori Hattori

Human Activity Recognizer for Mobile Devices with Multiple Sensors

This paper presents a novel human activity recognizer used to estimate a users activities with sensors on a widespread consumer mobile device. Our recognizer can estimate a users means of migration by using a combination of an acceleration sensor and a GPS. We evaluate the accuracy rate of the estimation with cellular phones carried freely and the result is 90.6 percent even in the case of intermittent working mode for lower power consumption.

Agent-based drivers' information assistance system

Abstract“Drivers’ Information Assistance System (DIA system)” is an ITS (Intelligent Transport Systems) application framework that provides agent-based information assistance to drivers through car navigation systems or on-board PCs.DIA system enables flexible information retrieval over the Internet using intelligent mobile agent, and incorporates a high-speed event delivery facility that makes real-time information service possible. The goal of the system is to provide up to the minute information and services related to driver needs, such as parking lot vacancy information. Crucial to making this a practical operation is the agent-based ability to access the network while the vehicle is in motion.

picoPlangent: an intelligent mobile agent system for ubiquitous computing

This paper describes an intelligent mobile agent named picoPlangent that we developed for use with portable devices. picoPlangent is designed with a component-based architecture. The agent functions are implemented by a set of small components, and the arrangement of these components can be easily changed within the limits of the available resources of each portable device. Agent actions are described into the goal tree that realizes the flexible actions of the agent on portable devices. The picoPlangent architecture is simple and easy to implement on various devices. We implemented picoPlangent using J2SE on PCs, GCC on PDAs (Palm devices) and J2ME on cellular phones (iAppli/ezPlus).

BDI Model-Based Crowd Simulation

We present a new approach for crowd simulation in which a BDI (Belief-Desire-Intention) model is introduced that makes it possible for a character in a simulated environment to work adaptively. Our approach allows the character to realize realistic behavior by adapting its action with the sensed information in a changing environment. We implemented a demo system simulating the BDI model-based NPCs that extinguishes a forest fire with a 3D game engine, Source Engine. We measured the performance to evaluate the scalability and the bottleneck of the system.

Character-Based LSTM-CRF with Radical-Level Features for Chinese Named Entity Recognition

State-of-the-art systems of Chinese Named Entity Recognition (CNER) require large amounts of hand-crafted features and domain-specific knowledge to achieve high performance. In this paper, we apply a bidirectional LSTM-CRF neural network that utilizes both character-level and radical-level representations. We are the first to use character-based BLSTM-CRF neural architecture for CNER. By contrasting the results of different variants of LSTM blocks, we find the most suitable LSTM block for CNER. We are also the first to investigate Chinese radical-level representations in BLSTM-CRF architecture and get better performance without carefully designed features. We evaluate our system on the third SIGHAN Bakeoff MSRA data set for simplfied CNER task and achieve state-of-the-art performance 90.95% F1.

Designing mobile search interface with query term extraction

Web searching with mobile phones is still not as easy as with Personal Computers (PCs). To solve the problem, we proposed a one-touch search application using first query term extraction from browsed Web pages [1]. This paper reports on a preliminary experiment that we conducted with 50 subjects in order to design the interface of the application. The results of the experiment indicate that although providing query term candidates to users reduces the cost of query input, the complexity of the query input process poses a problem.

First query term extraction from current webpage for mobile applications

Inputting query terms on a mobile terminal is frustrating because of device limitations though mobile web search is becoming popular. Query prediction is a promising approach for mobile search. In the literature, however, little attention has been paid to the first query term, though there are many reports on query expansion or second query recommendation. In this paper, we propose a first query prediction method that enables users to search related information for the current browsed webpage with an easier interface such as touch operation without any background search processes. The proposed method consists of body-text extraction, candidate query term extraction, and a scoring process. We also implemented a one-touch search application that works for Japanese webpages at a practical speed on a smartphone. According to our closed evaluation with 299 webpages and open evaluation with 14 users, our method achieved practical quality.

Planning mobile agent architecture for open network environment

This paper describes a planning mobile agent architecture proposed for open network environments such as the Internet. Mobile agents in this architecture have mobile, planning, and meta-level control functionality, and address the locality, variability, and diversity problems to which open network environments are subject.

Acceleration Noise Correction for Transfer Inference Using Accelerometers on Mobile Devices

The acceleration noise generated in the ordinary usage of mobile devices (e.g. when “taking out” the devices or operating them) interferes with estimation of the means of migration using accelerometers on the devices. We developed a correction method for the noise generated when users take out devices, change their posture, and operate the devices. The method uses the changes of acceleration and the operation events acquired from the operating systems of the mobile devices to detect the period of noises. The result of evaluation shows that the method using the acceleration changes improves the precision of the context inference approximately 5 %, and the method using the operation events corrects the inference mistaking resting for boarding.

Semi-automatic Evaluation System for Supporting Term Extraction Application Development

Important term extraction from the current web page is a common process in many client-based web applications such as query-free search, search query recommendation, information extraction, user profiling, and personalization. However, the tuning cost is large for deciding parameters when updating modules or dictionaries, or when porting an application for other languages or devices. We propose an evaluation system to collect web pages and corresponding correct data with visual check and automatic tuning of parameters. We confirmed the practicality of this system through the development and the two-month refinement of first query term extraction module.