一种基于语音识别的室内定位方法

NewApproach for Indoor Localization Based on Speech Recognition

基于位置的服务在生活中扮演着越来越重要的作用,不仅可以为用户提供定位导航服务,而且改变着人们的生活交流方式.现有的室内定位系统大都是利用Wi-Fi接收信号强度(RSS)进行指纹匹配,以坐标方式表示位置,存在着定位精度低,不能直观表示室内位置关系等诸多不足.针对这些不足,本文提出一种基于语音识别的室内定位方法-HspotNavi.当用户静止时,利用语音识别和条件随机场解析周围声音中的位置语义,并将其与用户行走过程中采集的WiFi指纹序列、惯导数据一起打包成路径信息,上传至服务器.当其他用户请求定位服务时,HspotNavi根据语义获取预存路径,利用Wi Fi指纹序列的错位匹配提供定位服务.

Service based on localization is one of the most important contexts for human activities, and it can not only provide a service to locate or navigate, but changes the way of our life and communication. In recent years, the Wi-Fi fingerprint （ WF ） technique has been actively studied, but in the most of existing WF-based system, the RSS is directly adopted as fingerprint, which may cause a significant performance degradation. Moreover, the location is generally showed in the coordinate system, which can not intuitively reflect the relationship of indoor position. To tackle the aforementioned problems, this paper presents a speech recognition-based localization solution--HspotNavi. The proposed solution extracts the semantics feature for space partition via Automatic Speech Recogni- tion and CRFs, and samples WiFi and IMU sensors as guiders walk along pathways to destinations. HspotNavi automatically processes all sensor data and packs them into a tracjectory, which can be shared via a cloud server. When other users send a localization requirement, HspotNavi adopts dislocations matching of the coincident Wi-Fi fingerprint sequence between prestored tracjectory and current tracjectory to provide localization service.