| Automobile brings great convenience to people’s life,but also brings threat to people’s personal safety.Fatigue driving is one of the most likely causes of traffic accidents.Therefore,timely and efficient fatigue driving detection and early warning of relevant personnel is particularly important.Fatigue driving detection has become a hot research topic.Fatigue driving detection,according to the different data acquisition equipment,can be divided into contact detection and non-contact detection.Contact detection mainly uses wearable devices for personal detection,which has high accuracy,but is limited by the endurance of the battery.The non-contact detection mainly uses camera equipment for data acquisition,which has higher accuracy and is more convenient to use.However,it has higher requirements for lighting conditions.At the same time,image acquisition also brings the problem of privacy leakage.These two traditional detection methods need to purchase additional equipment,so we need a more new lightweight way to detect fatigue driving behavior.Using audio sensing technology for fatigue driving detection is a new hot spot in the field of non-contact detection.It has the advantages of high accuracy,good concealment,no privacy leakage and so on.At the same time,it can use the smart phones,bracelets and other devices to achieve the audio collection,without wearing it when carrying out relevant detection.It is an advanced technology that can be perceived without touching.This paper proposes a fatigue driving detection model based on audio perception.The specific research content is as follows:(1)A blink frequency estimation algorithm based on audio is proposed.The collected acoustic signal is preprocessed,and the accurate blink frequency is obtained by local extremum method after noise reduction.The obtained blink frequency can be used as a parameter to judge fatigue state.(2)A multi feature fatigue state recognition algorithm model based on deep learning is proposed.With the help of long-term memory network and deep neural network,a fatigue driving detection system based on blinking,nodding,yawning and other features is realized.In order to avoid the error caused by individual differences in the blink based single feature fatigue driving detection scheme,multiple features are used to judge whether the driver is in fatigue state.(3)On the basis of theoretical research,a fatigue driving detection system is designed and implemented.A prototype system of fatigue driving detection is built on Android platform by using acoustic signal.With the help of the built-in speaker and microphone of the mobile phone,the audio signal is played and collected.Without additional hardware support,the smart phone is transformed into an active sonar sensing system,and the real-time detection of fatigue driving behavior is realized.The experimental results show that the average accuracy of the system is 94.1%,which can effectively detect the driver’s fatigue state.In conclusion,the fatigue driving detection technology proposed in this paper can effectively remind drivers to pay attention to driving safety,which is of great significance to ensure traffic safety. |
