What is it about?
This study presents a method for recognizing human emotions using LogNNet neural network and keystroke dynamics dataset. Two types of training sets were investigated, with 10 and 15 features compiled on the basis of the Emosurv database. It is shown that the accuracy of recognition of one emotion out of 5 (happy, sad, angry, calm, neutral state) reaches 33.4% when using 10 features read only from the keyboard.
Featured Image
Photo by Tengyart on Unsplash
Why is it important?
Determining the emotional state of a person in modern society acquires an important role when working in a factory or office. For example, in order to increase labor productivity, it is important to assess the fatigue and stress of workers and provide psychological assistance in time. Evaluation of the emotional background of adolescents in educational institutions can help prevent conflict situations and improve the level of education. Using the keyboard to determine emotions has its advantage, since the keyboard is a common and inexpensive instrument, and the development of this technique is the goal of this research.
Perspectives
Based on the presented results, it is possible to propose the concept of a device for assessing the emotional state of a person, which will be built into the keyboard, without communication with an external computer. With this approach, there is no need to install additional software on the computer. On such a keyboard, alarm can display when a specific emotion is recognized, for example, anger. Such devices can be in demand for dispatchers, call center operators, where the quality of the services provided depends on the emotional state, and in educational institutions to control the emotional background. All models under study can be placed on microcontrollers with RAM up to 3 kB, which is a weighty argument for applying the described methodology for edge computing in IoT.
Dr. Andrei Velichko
Petrozavodsk State University
Read the Original
This page is a summary of: Emotions recognizing using lognnet neural network and Keystroke dynamics dataset, January 2023, American Institute of Physics,
DOI: 10.1063/5.0162572.
You can read the full text:
Contributors
The following have contributed to this page
