| 摘要 |
<p>The evaluation of alertness of an automobile driver to analyze eyelid movements, comprises associating every blinking with a duration vector of each component, calculating a cumulated duration vector of each component having summation of average duration and long duration blinking by a temporal analysis window, and deducing the comparison of the calculated values of average and long duration blinking with threshold values. The evaluation of alertness of an automobile driver to analyze eyelid movements, comprises associating every blinking with a duration vector of each component, calculating a cumulated duration vector of each component having summation of average duration and long duration blinking by a temporal analysis window, deducing the comparison of the calculated values of average and long duration blinking with threshold values, intergrating environmental information for introducing ponderation levels with a diagnostic data, which is based on the analysis of the eyelid movements, integrating the informations obtained from behavioral observations such as observation of the trajectory of the vehicle, validating the data of analysis window, if the number of blinking detected during the analysis is higher than the given threshold, and triggering the opening of the analysis window during every detection of blinking, where each opened analysis window covers a determined period of time before triggering. In the evaluation process, a preliminary stage comprises classifying the blinking duration of the drivers by implementing a fuzzy logic mathematical method, and classifying alertness of the drivers by N classes, where N is >=2. The blinking duration comprises three classes such as short duration blinking, average duration blinking and long duration blinking in which the classes are adapted to have border zone of progressive transition. The classification of alertness comprises four classes such as a waked up class defined by a number of average duration blinking lower to a given threshold, a first intermediate class corresponds to a little somnolence state defined by a number of average duration blinking higher to a given threshold and lower to an intermediate threshold, a second intermediate class corresponds to a somnolence state defined by a number of average duration blinking higher to intermediate threshold and by a number of long duration blinking lower to a given threshold, and a sleeping class defined by a number of long duration blinking higher to a given threshold. The little somnolence state is also defined by a number of long duration blinking lower to the given threshold. Analyzing the eyelid movements of the driver is determined by confidence degree of the blinkings. The confidence degree of the blinkings are determined by comparing two signals that are representative of the eyelid movements based on the criteria of predetermined comparison such as criteria relating to simultaneity, amplitude or gradients of the signals. The confidence degree (C 0>) is determined by the summation of blinking degree of each duration class multiplied by the confidence degree (Ci) of the blinking and the whole divided by the summation of blinking degree of each duration class. Combination of criteria relating to simultaneity, amplitude and gradients of two signals, which are representative of the two eyelid movements are adapted to determine the confidence degree associated with every blinking. The environmental information is a representative of conditions of control such as time of control, temperature in the vehicle, hour of the day, type of roadway (road, motorway, urban), and data concerning the driver such as age and experience.</p> |
