Teppei Nakano

Evaluation of Collaborative Video Surveillance Platform: Prototype Development of Abandoned Object Detection

This paper evaluates a new video surveillance platform presented in a previous study, through an abandoned object detection task. The proposed platform has a function of automated detection and alerting, which is still a big challenge for a machine algorithm due to its recall-precision tradeoff problem. To achieve both high recall and high precision simultaneously, a hybrid approach using crowdsourcing after image analysis is proposed. This approach, however, is still not clear about what extent it can improve detection accuracy and raise quicker alerts. In this paper, the experiment is conducted for abandoned object detection, as one of the most common surveillance tasks. The results show that detection accuracy was improved from 50% (without crowdsourcing) to stable 95-100% (with crowdsourcing) by majority vote of 7 crowdworkers for each task. In contrast, alert time issue still remains open to further discussion since at least 7+ minutes are required to get the best performance.

Speech Interface Evaluation on Car Navigation System - Many Undesirable Utterances and Sever Noisy Speech -

Recently, ASR (Automatic Speech Recognition) functions have commercially been used for various consumer applications including car navigation systems. However, many technical and usability problems still exist before ASR applications are on real business use. Our goal is to make ASR technologies for a real business use. To do so, we first evaluate a car navigation interface which has ASR as an input method, and second evaluate an ASR module using real noisy in-car speech. For ASR applications, we envision mobile environments, e.g. mobile information service systems such as car navigation systems and cellular phones on which an embedded speech recognizer (Kokubo et. al., 2006) is running and which are connected to remote servers that support various information-seeking tasks. Taking a look at commercially available car navigation systems, currently over 75% systems have ASR interfaces, however, there are very few drivers who have experiences to use the ASR interfaces. What is the problem? This is caused by the ASR usability problems. In this chapter, we report two experimental evaluation results of ASR interface for mobile use, especially for car navigation applications. First, we evaluate the usability aspects of speech interface and second, we evaluate in-car noise speech problems to propose an effective method to cope with noisy speech. For the first evaluation, we use a prototype which has a promising speech interface called FlexibleShortcuts and Select&Voice produced by Waseda University (Nakano et. al., 2007). We found many undesirable OOV (Out-OfVocabulary) utterances which make the interface worse. From the second experiment to check car-noise problems, we propose an array microphone + Spectrum Subtraction (SS) technique to increase recognition accuracy.