Problem statement:- The increasing usage of speech recognition systems by people has led to the ease of access in their day to day lives. People use personal assistants like Apple’s Siri, Microsoft Cortana or Google Now and make their lives easier, however people with speech impairments like stuttering cannot benefit from these services because these companies have catered their speech recognition algorithms to the majority of people, that is, the people without any speech disorders even when about 70 million people in the world suffer from stuttering alone. These voice recognition systems are unable to detect when people afflicted by stuttering use it because when the person starts stuttering the service thinks that the person has completed speaking and doesn’t process what comes subsequently.

Solution:- In order to make these above applications more universal, we have developed a platform to recognize, remove stutter (repetition of words and long pause) and convert an audio of speech into recognizable speech. We have first removed noise from the audio input then the acoustic and pitch related features such as energy and correlation factor are used to detect and remove repetition and long pause. We will be implementing the above solution in a search engine. We have used Bing API provided and modified it to include voice search.

We have used MATLAB to remove stutter from an audio input, besides this we have used Node JS and JavaScript for implementing this in Bing Search Engine.