Research On Early Warning Of Construction Workers’ Work Safety Based On Speech Emotion Recognition

Although the number of safety accidents that have occurred in the field of construction engineering in my country in the past ten years has been declining year by year,the problem of engineering safety is still a problem that requires special attention.At this stage,my country’s economic and social development has entered the stage of relying on innovation.Similarly,more new technologies should be adopted in the field of construction engineering to lead the industrial upgrading of the construction industry.This paper applies machine learning to safety early warning research in construction engineering,uses convolutional neural network models to identify the emotions of construction workers,and quantitatively analyzes the emotions of construction workers during operation,and builds a safety early warning model.Early warning signs in emotional states.Based on the existing research at home and abroad,this paper first combs the related research results in speech emotion recognition,including speech emotion database,speech emotion characteristics,speech emotion recognition algorithm and related research on the application of speech emotion recognition,and optimizes it The shortcomings are analyzed and reviewed.Secondly,it analyzes the influence of human emotions on human behavior based on psychology related theories,accident cause theory,social cognitive theory,emotional event theory and field dynamics theory.After that,I sorted out the safety performance related research in the field of construction engineering,and then analyzed the correlation between the emotions of construction workers during operation and the occurrence of safety accidents through questionnaire surveys.Then the neural network is trained by preprocessing the voice signal in the ACCorpus voice emotion database,extracting the emotional feature of the voice signal,and inputting the extracted features into the convolutional neural network to finally obtain a volume with certain voice emotion recognition capabilities on the database Product neural network.In addition,by reconstructing the speech emotion database,improving the parameter update method of the convolutional neural network,the setting method of the initial value of the weight,and suppressing over-fitting to improve the recognition rate of the speech emotion recognition convolutional neural network and prevent over-fitting to improve the model The generalization ability.Finally,the advantages and disadvantages of existing safety early warning methods are analyzed,and a model that uses the emotional fluctuation value of construction workers to carry out related early warnings is constructed,and the management measures that managers can take after the early warning signals are issued are proposed.