A Novel Idea for Designing a Speech Recognition System Using Computer Vision Object Detection Techniques

Abstract

It is very challenging for establishing the communication with deaf people around the world. They need to get the assistance from others, and others need not be true always. To overcome this situation, a device we propose to develop an application which will provide the easy way of communicating using sign language without the help of others. The concept of this device development is a novel idea. It is intended to make the device as standalone using the recent development in embedded system technology. The proposed system is aimed to develop a pocket assistant for the deaf and hearing-impaired people in communicating with other people. All the functionality of the application is built around the organization of communication to establish a conversation between the user and his interlocutor. It is planned to develop a device to recognize the interlocutor’s speech in real time, query the related sign representation stored in database, and display the text or set of pictures in sign language on the screen. The device will be based on Raspberry Pi hardware. The technology involves capturing the audio using …. The audio will be processed by removing the noise and fed into the audio-to-text convertor to output the text message. Text information is detected using histogram of oriented gradients (HOG) and local binary pattern (LBP). The required information will be selected and queried to extract the desired sign representation from the database and provide the desired output to the user on the screen. Technological transfer of the proposed product will enable mass production that can be utilized in national and global market for the benefit of the elderly and deaf people. It has three modules, login, recording the information, and translating the information and storing it in the database.