Abstract
Fingerprinting acoustic localization usually requires tremendous time and effort for database construction in sampling phase and reference points (RPs) matching in positioning phase. To improve the efficiency of this acoustic localization process, an iterative interpolation method is proposed to reduce the initial RPs needed for the required positioning accuracy by generating virtual RPs in positioning phase. Meanwhile, a two-stage matching method based on cluster analysis is proposed for computation reduction of RPs matching. Results reported show that, on the premise of ensuring positioning accuracy, two-stage matching method based on feature clustering partition can reduce the average RPs matching amount to 30.14% of the global linear matching method taken. Meanwhile, the iterative interpolation method can guarantee the positioning accuracy with only 27.77% initial RPs of the traditional method needed.
Highlights
With the development of signal processing technology and artificial intelligence technology, voice interaction has been gaining extensive attention in the smart device field [1,2,3]
In the online positioning process, the virtual reference points (RPs) can be generated by the iterative interpolation method, as Figure 4 shows, where the iteration interpolation process is based on four adjacent RPs
The positioning database consisting of 72 initial RPs is divided into four sub-databases by K-Means clustering algorithm, and four adjacent RPs selected by the two-stage matching method are used for 13 test point’s location estimation based on iterative interpolation
Summary
With the development of signal processing technology and artificial intelligence technology, voice interaction has been gaining extensive attention in the smart device field [1,2,3]. Most existing acoustic localization methods are parametric positioning methods, which are based on the space geometrical propagation models of acoustical signal [6,7,8,9,10,11].
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
