Measurement of signal use and vehicle turns as indication of driver cognition

Abstract

This paper uses data analytics to provide a method for the measurement of a key driving task, turn signal usage as a measure of an automatic over-learned cognitive function drivers. The paper augments previously reported more complex executive function cognition measures by proposing an algorithm that analyzes dashboard video to detect turn indicator use with 100% accuracy without any false positives. The paper proposes two algorithms that determine the actual turns made on a trip. The first through analysis of GPS location traces for the vehicle, locating 73% of the turns made with a very low false positive rate of 3%. A second algorithm uses GIS tools to retroactively create turn by turn directions. Fusion of GIS and GPS information raises performance to 77%. The paper presents the algorithm required to measure signal use for actual turns by realigning the 0.2Hz GPS data, 30fps video and GIS turn events. The result is a measure that can be tracked over time and changes in the driver's performance can result in alerts to the driver, caregivers or clinicians as indication of cognitive change. A lack of decline can also be shared as reassurance.

An Integrated Approach for Episodic Cognition Assessment

“…Driving is a very complex cognitive task including executive function tasks such as trip and route planning through to highly trained and learned tasks such as turn signal use. The measurement of driving is presented in three sections with an initial focus of measurement of driving behaviours that are unique to the driver, the measurement of navigational performance as an indication of cognitive ability and the measurement of turn signal use as one measure of vehicle operation (refereed journals [10] and conferences [12][13][14][15][16][17][18][19][20][21]).…”

Reliable, Real-Time Acceleration Measurement Toward Automated Monitoring of Driver Behavior

“…Gyllensten et al, however, have shown that these accuracies can be misleading, as movements in the real world are rarely as clearly defined as they are in the laboratory setting [58]. [64]. These studies have shown that sensor equipment is capable of providing a clearer understanding of drivers'…”

Characterization of Stable-Health Older Drivers Using Low-Speed Driving Maneuvers From In-Vehicle Sensor Data

“…as changes in heading of 70 o to 110 o across the entire event (i.e., total change in heading from the start to end of the event). Wallace et al [56] identified when the vehicle was turning using only the GPS data from a smartphone GPS sensor and the proposed turnidentification algorithm was based on the change in heading of 30 o between sampled GPS coordinates; their performance was 73% true positive rate, 3% false positive rate, and 14% false negative rate tested on 215 true turns. In the same study, the turn-identification algorithm was compared to using the Google Maps GIS algorithm for identifying turns.…”

“…The three turn-identification algorithms involved locating the coordinates that occur within a turn, and then locating the start and the end of the turns. The first algorithm was based on a threshold on the change in heading to find significant corners and bends, and this algorithm was implemented by Wallace et al [56] for counting the number of turns in a trip. The second algorithm was an extension of the first algorithm by simplifying the GPS trace before applying a threshold on the change in heading to emphasize the corners.…”

“…Since there may be thousands of trips taken by a driver, a computer algorithm must be implemented to automatically and effectively identify the corners and bends in the GPS trace. Wallace et al [56] have studied the turning maneuvers from GPS data and they used a common approach by applying a threshold to the change in heading to filter for turning points, which are the coordinates that lie within the turn itself. The proposed algorithm by Wallace et al [56] had a 73% true positive rate, 3% false positive rate, 14% false negative rate for a threshold of 30 o in change in heading.…”