Volume 10 Issue 3 - May 2018

  • 1. Real-time implementation for digit recognition using raspberry pi

    Authors : Snehal R. Kalbhor, Ashwini M. Deshpande

    Pages : 50-56

    DOI : http://dx.doi.org/10.21172/1.103.09

    Keywords : Contour, Convex hull, K-NN, LDA-LR, PCA-LR, SVM,OpenCV, Raspberry Pi

    Abstract :

    The identification of sign language is a bridge between the normal people and people with hearing and speech problems. Most of the people are unable to recognize the gestures made by these people. Computer vision and machine learning techniques can be used to remove this barrier by creating a working model which automatically detect the gesture performed by them. In this paper a real-time implementation for digit recognition based on the Raspberry Pi with camera module is presented. Raspberry Pi is programmed with Python programming language supported with OpenCV library. In this system template matching, Linear Discriminant Analysis-logistic Regression (LDA-LR), Principal Component Analysis Logistic regression (PCA-LR), K-Nearest Neighbors (K-NN) and Support Vector Machine (SVM) learning based methods areused for digit recognition and their performance is compared. In template matching method, the template of the gesture is matched with the database image by using correlation function while in LDA-LR, PCA-LR, KNN and SVM based systems,the distance between the palm centroid and fingertip is considered as a feature and the features are trained with LR, KNN and SVM Models.The proposed system achieved the accuracy of 63.76% for LDA-LR, 72.46% for PCA-LR, 77.77% for K-NN and 83.09% for SVM approach.Template matching algorithm shows 100% matching result only if it is compared with exactly same template.

    Citing this Journal Article :

    Snehal R. Kalbhor, Ashwini M. Deshpande, "Real-time implementation for digit recognition using raspberry pi", https://www.ijltet.org/journal_details.php?id=932&j_id=4599, Volume 10 Issue 3 - May 2018, 50-56, #ijltetorg