abstract
- Globally the improvement and evaluation of the academic performance of students has been a priority, however the way in which the quality of learning process is evaluated within the classroom is based on traditional methods such as grades or perception surveys. Additionally, measuring continuously the performance of the student, teacher and its interaction in the classroom is difficult because there are several internal and external factors that can affect the pedagogical practice in the classroom or e-learning environments, and currently, their effects are not completely understood. Currently, advances in motion tracking through low cost devices such RGBD cameras allows the real-time monitoring of persons posture inside closed spaces such a classroom. Some research projects have associated posture with affective and cognitive state, but as far as we know none have proposed an approach to classify learning and teacher interaction using posture. An approach that uses a set of performance metrics of the student and teacher, in order to classify whether learning and teacher-student interaction was successful is developed and tested. This was an experimental design using an experimental and control group in order to evaluate if it is possible to classify between poor and good interaction between teacher and student. The results showed that it is possible to classify between poor and good interaction between teacher and student, besides the best method of classification is the approach based on neural networks with an accuracy of 76%. The proposed approach could classify whether an interaction between the student and the teacher was good or not. The results showed that the best method of classification was the approach based on neural networks with an accuracy of 78%.