Posted by Admin: System Admin
As one of the biggest contributors to road accidents and fatalities, drink driving is worthy of significant research attention. However, most existing systems on detecting or preventing drink driving either require special hardware or require much effort from the user, making these systems inapplicable to continuous drink driving monitoring in a real driving environment. In this paper, we present DetectDUI, a contactless, non-invasive, real-time system that yields a relatively highly accurate drink driving monitoring by combining vital signs (heart rate and respiration rate) extracted from in-car WiFi system and driver’s psychomotor coordination through steering wheel operations. The framework consists of a series of signal processing algorithms for extracting clean and informative vital signs and psychomotor coordination, and integrate the two data streams using a self-attention convolutional neural network (i.e., C-Attention). In safe laboratory experiments with 15 participants, DetectDUI achieves drink driving detection accuracy of 96.6% and BAC predictions with an average mean error of 2 _ 5mg/dl. These promising results provide a highly encouraging case for continued development. Machine learning is an important component of the growing field of data science. Through the use of statistical methods, different type of algorithms is trained to make classifications or predictions, and to uncover key insights in this project. These insights subsequently drive decision making within applications and businesses, ideally impacting key growth metrics. Machine learning algorithms build a model based on this project data, known as training data, in order to make predictions or decisions without being explicitly programmed to do so. Machine learning algorithms are used in a wide variety of datasets, where it is difficult or unfeasible to develop conventional algorithms to perform the needed tasks.
A. Drunkenness Detection Hardware-Based Detection: First used in the United Kingdom in the 1970s [17], breathalyzers are the world’s most commonly used tools for testing inebriated drivers. Over its years of usage, researchers have connected breathalyzers, as well as other types of breath alcohol sensors, to smartphones via Bluetooth to improve BAC tracking, especially for self monitoring by drivers themselves. Example systems include: BACtrack Mobile Pro [18], Breathmeter [19]. One major disadvantage of breathalyzers is that the results are highly susceptible to the oral environment [20] and certain diseases (e.g., diabetes, liver and kidney diseases [20]), which may lead to false detection. Alternatives to breathalyzers include SCRAM, a transdermal sensor that measures the wearer’s BAC through their sweat every 30 minutes [21]. The same kind of system is available in a tight wristband that fits closely to the skin [4]. However, SCRAM-based systems require a close contact between the skin and the sensor. Any space or anything between the skin and the sensor will affect the detection accuracy. Moreover, these systems require users to purchase extra devices or sensors, which may be expensive. Camera-Based Detection: Camera-based drunk driving systems have also been developed [22], [23]. In [22], facial landmarks and motions are recognized in images to detect whether the driver is drunk driving or not. In [23], an audiovisual database is utilized to realize bimodal intoxication detection. However, camera-based approaches are sensitive to lighting conditions and there is potential risk of privacy violation [24]. Behavior-Based Detection: The side effects of alcohol consumption include arrhythmia [14], slowed respiratory rates [15], impaired psychomotor performance [8], and unsteady gait [6]. This abnormality in vital signs and behaviors can be leveraged to detect whether the user is under the influence of alcohol. Bae et al. [7] developed a smart phone based system to track the drinking episodes of users based on built-in sensors (e.g., accelerometer) and the smartphone status (e.g., battery and network usage). Leveraging alcohol’s influence on motor coordination and cognition, Markakis et al. [8] designed five human-computer interactions to detect BACs (such as swiping or touching the screen in particular ways), akin to the finger-to-nose DUI tests. However, these works require users to interact with their phones (swipe the phone or engage in games), which interrupts the driving task and cannot offer a continuous drunk driving detection. Disadvantages ? An existing methodology doesn’t implement variational mode decomposition method. ? DetectDUI can't measure a person’s vital signs through WiFi signals and their psychomotor coordination through steering wheel operations.
• As far as we are concerned, DetectDUI is the first contactless method of detecting drink driving, including measuring the driver’s BAC that can be administered while driving. • We have proposed a series of signal processing algorithms for extracting human vital signs from WiFi signals given chest motions with high levels of accuracy. • We have proposed to use C-Attention to combine the information of vital signs and psychomotor coordination to reach a well-round drink driving prediction. • Extensive experiments on 15 individuals show DetectDUI is able to distinguish normal driving from drink driving in real-time with a 96.6%-accurate estimation and the driver’s BAC to within an MAE of 0.002% to 0.005%. Advantages ? The proposed system DetectDUI detects drink driving and predicts BAC through a driver’s vital signs and psychomotor coordination.The system shows the architecture of DetectDUI. In DetectDUI, vital signs are tracked through a WiFi sensing system and writing as datasets. ? The system proposes a novel adaptive variational mode decomposition (AVMD) method to separate the mixed signal into multiple modes, and then keep the modes that relate to breathing and heartbeat respectively.