Use of KNN Classifier for Emotion Recognition Based on Distance Measures
Vishal D. Bharate1, Devendra S. Chaudhari2, Mayur D. Chaudhari3
1Vishal D. Bharate, Department of Electronics and Telecommunication, Government College of Engineering, Amravati/ Sinhgad Academy of Engineering, Pune, India.
2Devendra S. Chaudhari, Department of Electronics and Telecommunication, Government College of Engineering, Jalgaon, Jalgaon, India.
3Mayur D. Chaudhari, Data Architect, Parkar Labs, Pune, India.
Manuscript received on October 01, 2019. | Revised Manuscript received on October 15, 2019. | Manuscript published on October 30, 2019. | PP: 1294-1298 | Volume-9 Issue-1, October 2019. | Retrieval Number: A9639109119/2019©BEIESP | DOI: 10.35940/ijeat.A9639.109119
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Human-computer interaction (HCI), in recent times, is gaining a lot of significance. The systems based on HCI have been designed for recognizing different facial expressions. The application areas for face recognition include robotics, safety, and surveillance system. The emotions so captured aid in predicting future actions in addition to providing valuable information. Fear, neutral, sad, surprise, happy are the categories of primary emotions. From the database of still images, certain features can be obtained using Gabor Filter (GF) and Histogram of Oriented Gradient (HOG). These two techniques are being used while extracting features for getting the expressions from the face. This paper focuses on the customized classification of GF and HOG using the KNN classifier.GF provides texture features whereas HOG finds applications for images exhibiting differing lighting conditions. Simplicity and linearity of KNN classifier appeals for its use in the present application. The paper also elaborates various distances used in KNN classifiers like city-block, Euclidean and correlation distance. This paper uses Matlab implementation of GF, HOG and KNN for extracting the required features and classification, respectively. Results exhibit that the accuracy of city- block distance is more.
Keywords: Gabor Filter, Histogram of Oriented Gradient, Human-computer interaction, k-nearest neighbors.