Research On Driver Road Rage Emotion Recognition Based On Multi-feature Fusion

Road rage is a potential factor for driving safety.The driver emotion state monitoring and danger in time warning can effectively reduce traffic accidents.At present,the reliability of driver emotion recognition based on single factor for road rage needs to be further improved.Therefore,based on multi-feature fusion,this research proposed a recognition method of driver road rage emotion state by fusing facial expression,speech and text features.Then proposed road rage recognition Support Vector Classification(SVC)model verified by experiments.The main research contents of this research include:(1)Feature extraction and database construction of road rage facial expressionFacial expression change is a typical reflection for road rage.Road rage can be recognized is feasible based on changes of facial expression.Due to the facial features extraction is easily affected by objective factors such as image zooming,transforming,rotation and face geometric differences,which make the driver emotion recognition very difficult.Therefore,based on the analysis of the relationship between facial features and driver emotion,this research proposed flexible template set for road rage feature extraction.Based on these template parameters,the driver can be recognized reliably.(2)Feature extraction and database construction of road rage speech and textSpeech and text are important information that can reflect driver road rage,and road rage state recognition based on speech and text features is effective.Due to the lack of in-depth analysis and comprehensive extraction of speech feature elements and the high redundancy of text feature,the recognition reliability is poor.Therefore,based on the in-depth analysis of the correlation between speech features and driver emotion changes,this research comprehensively extracted the key features of prosodic features,tone quality features and spectral features of speech based on the constructed driver speech and text data set.Secondly,a text feature extraction method based on chi-square statistics and TF-IDF is proposed to realize the effective recognition of road rage emotion state.(3)Construction and demonstration of road rage emotion recognition model based on SVCBased on the above characteristics,the key to improve the recognition and classification accuracy is to select the appropriate classification model for road rage recognition.Because the SVC model used for classification in Support Vector Machine(SVM)has a good classification effect on small sample data.Therefore,this research proposed to use cross validation method to build a SVC model based on facial expression features,speech text features and fusion of facial expression,speech text features for road rage recognition,and combined with relevant experimental data to carry out empirical analysis of the three models.Aiming at the problem of road rage recognition,this research pays more attention to the reliability of road rage analysis and the method of state recognition.In which,related breakthroughs in this research are as the followings:Aiming at the lack of relevance of facial features to road rage emotion,this research constructs a set of facial apparent elastic template changes,and extracts key features of road rage based on the template parameter change rules,which effectively solves the problem of road rage emotion recognition based on facial expressions.Aiming at the insufficient reliability problem of road rage representation caused by single consistent factors,this research adopts the multi-fusion method of fusing speech text features and facial appearance features for road rage emotion state recognition,thus effectively improving the accuracy of road rage emotion state recognition.