In this project, it is aimed that school counselors can accurately and objectively perceive the emotional state of students who do not show their emotional state, and eliminate negative situations in students' lives with positive psychology interventions and increase their social and academic success.
This project has been developed to bring an innovation in the field of AI-based emotion status analysis. Emotion analysis is the process of identifying and analyzing emotions using image and audio data. The emotions of the students were analyzed based on the voices and facial expressions of the students. There are 7 different emotions in the project. These emotions are happy, sad, angry, surprise, disgust, neutral and scared. At the end of both projects, the results of the students' emotions were taken correctly, and the results of both projects were compared and the success rates were calculated.
The proposed project is divided into two main parts, they are Emotion analysis and speech recognition. In the project, a number of technologies and methods were used to analyze people's emotional states and the words they use. With the selected CNN method, it worked for us to capture features (features) in different images (moments in the video) and classify them. On the other hand, the speech recognition part is an innovative development in the way people interact with devices. Thanks to speech recognition, we have been able to convert audio recordings of students whose emotional states we cannot understand into a "txt" file and then perform emotion analysis from each sentence. Also in this section, in order to avoid the problem of accents, "recognize.google" was used. Regarding another sound, we have taken precautions against minimum decibels and excessive , allowing users to choose the best experience to show by perceiving their behavior and sound in a clean and accurate way.
The main purpose of this project is to evaluate the emotional state of students as objectively and accurately as possible, to improve the success and quality of students in their educational and social lives. When the application is successfully performed, many School counselors will be able to improve the psychology of their students in a positive way thanks to this software.
Our success criteria in our project are as follows;
- Detection of faces and facial expressions with high resolution. 60% detection accuracy.
- Clean detection of minimum ambient noise.
- Using temporal information to better detect and track faces and facial expressions.
- To clearly identify and understand the words used.
In the work for the future, a more efficient algorithm can be used. In addition, if a data set is created based on the facial expressions of people in the country, these studies can give faster and more accurate results. In addition, our study is currently only used in the field of education, and it can be used in other fields in the future.
#Footnote: fer2013 dataset was used in this study.