Research On High-Precision Indoor And Outdoor Scene Recognition Technology Based On Multimodal Fusion

Now mobile phone has become an important platform of context-aware and communication. The perception and communication module of mobile phone can provide positioning and context-aware services. However, they have different applicability and performance and vary under different scenarios. The indoor/outdoor detection can provide essential and primitive information for upper-layer mobile applications. For example, we can take advantage of it to decide whether to turn on GPS and WLAN or not. While practically useful, the problem of indoor/outdoor detection has not been thoroughly studied yet. As we all know, detecting indoor/outdoor environments solely with GPS can be slow and inefficient. There are some other works relying on dedicated devices to assist the ambient environment detection. The deployment cost of such infrastructure-based approaches significantly limits the flexibility and scalability for general purpose detection. On the other hand, some recent works study the problem of logical localization by sensing the surrounding environment. Such an approach is unlikely to be generalized to deal with universal indoor/outdoor detection. Many works in image processing and pattern recognition study the problem of the indoor/outdoor image classification and automatic image tagging. Such approaches cannot directly be applied to our problem, since they require explicit, manual input from users. Considering the problem above, we will study a kind of high-precision indoor and outdoor scene recognition technology based on multimodal fusion.In this paper, we provide a kind of indoor and outdoor scene recognition technology:a lightweight sensing service which runs on the mobile phone and detects the indoor/outdoor environment in a fast, accurate, and efficient manner. We utilize the build-in sensor of mobile phone such as light sensor, geomagnetism sensor, temperature sensor, pressure sensor, acceleration sensor, and base station signal. Through data mining, modeling and experimental analysis, we find they show different features in the indoor and outdoor scenes and can be used to classify the surrounding environment. We prototype the system on Android mobile phones based on multimodal fusion and evaluate the system comprehensively with data collected from20traces which include84different places during two months period, employing different phone models. Ultimately, we obtain the result of accuracy more than90%,3seconds time delay and average power consumption of5mw.