Using Keyboard Handwriting for Control in Learning Systems

The problem of student control with the help of keyboard handwriting is considered in the article. While working in an educational environment (for example, during certification), one of the users can perform tests for another. In this case, it is important to determine who exactly works at a particular moment in the training system. It is necessary to monitor continuously. It is advisable not to use personnel for this, as this is an unproductive waste of time. We assume that the full list of people working with the program at the moment is known. For analysis, data on the keyboard handwriting of each student are used, the characteristics of these data are determined in advance. The available information is represented as a set of vectors. Each of them contains a calculated handwriting characteristic corresponding to a particular user. The decision about the identity of the user is made depending on the similarity of the current and saved handwriting. To do this, the characteristics are compared at the present time with the saved ones. The scalar and vector methods of forming admissible areas for handwriting are considered. In the first case, a spherical region is formed, which simultaneously takes into account all the characteristics. The radius is defined as half of the minimum distance between points, which are vector coordinates that correspond to the characteristics of the keyboard handwriting. In the vector case, several linear regions are defined. Each area corresponds to one characteristic of the keyboard handwriting. The size of the area depends on the distance between the corresponding coordinates of the vector containing the data about the keyboard handwriting. Experimental data are presented to test the performance of the proposed control method. Keyboard handwriting contains the average values and standard deviations of the durations of pressing the keys and intervals of pauses. It is shown that in the scalar case the resulting regions have a larger radius than in the vector one. This reduces the number of type I errors.