Application And Research Of LOS And NLOS Acoustic Localization Signal Classification Based On SVM

Acoustic signal itself contains a lot of information,so it can be used flexibly and efficiently in many aspects after stabiling its own physical properties.Based on this,in recent years,there have been many systems based on specific frequency band acoustic signal and after complete the research and development project,it will achieve the established goals.Related technologies are also developing maturely and those developments based on related disciplines and interdisciplinary high-quality applications are constantly emerging.At the same time,it also shows the huge development prospects of acoustic signal itself from many aspects.In order to make this part of acoustic signal process can be used in the indoor localization system,it is necessary to complete the corresponding research for the core technology of acoustic localization signal.Related technologies include but are not limited to feature extraction algorithms,feature selection and recombination,acoustic signal classification part and build related evaluation index construction.In the indoor localization system based on specific frequency band acoustic signal with self-calibration anchor position,the key research center of acoustic signal is: in one NLOS environment,due to the scattering,reflection and irregular attenuation of acoustic signal,the original localization algorithm can not play its original effect.Therefore,the processing of acoustic signals in NLOS situation is necessary.Before that,the classification of acoustic localization signal also turns out to be necessary.According to this basis,the development should be completed based on the appropriate strategy.In the process of rearrangement and selection,based on results of acoustic signal classification part,some adjustments are made for the NLOS situation,so that the original localization algorithm used in the indoor localization system can play a better role.After actual measurement,the upgrading method for indoor localization system can enhance the positioning effect in the actual specific scene in the NLOS phenomenon compared with the original system indeed.In this paper,based on those situations,this paper will focus more on feature extraction and training upgrade,LOS and NLOS acoustic localization signal wav files classification part,core training process upgrade,acoustic signal eigenvalue regression analysis,complete the established research process and finally giving the real effect comparison of each optimization scheme under NLOS circumstances or conclustions of great value.The central work of this paper will include the following three aspects: first of all,according to the existing acoustic localization signal feature extraction algorithm,complete the upgrading of partial feature weight,highlight the feature function beneficial to classification,and carry out partial reorganization.In the actual situation,these results will be used as the input of the classifier,and more adjustments should be made according to the characteristics of NLOS acoustic signals,so as to highlight the discrimination of NLOS acoustic signals;Secondly,although the classical SVM classifier is still efficient,it can not perfectly realize NLOS acoustic localization signal identification easily.Therefore,it is necessary to introduce the improved swarm intelligence optimization strategy to complete the upgrading method from three aspects: feature rearrangement,dynamic adjustment of training weight and convergence speed of training process.Based on this,the development part,comparison part of different feature combinations and corresponding upgrading schemes are completed.The classification results will directly affect the actual process of indoor localization system in NLOS situations,and provide the possibility for of updating positioning accuracy in complex phenomenon;Then,in view of the problems of NLOS acoustic signal in some indoor localization systems,such as difficult to receive,difficult to obtain and strong randomness,the regression analysis part based on fuzzy information granulation model is also introduced to complete the regression analysis of NLOS acoustic signal,so that the classifier can obtain more useful training samples,enhance the actual effect of the classifier.Finally,the actual positioning accuracy as for indoor localization system will be strengthened.