TeachSpeech

TeachSpeech Logo and Cover page… Let’s Get Started!

TeachSpeech is a platform independent online web application for interactive speech therapy. It uses speech recognition and gives real-time feedback on the accuracy of one’s speech. This project is mainly aimed at people who lost their flexibility in speech due to any mishap or has a stuttering problem. This online & interactive application will be very helpful for those who are unable to undergo therapy with professional therapists due to budget problem or unavailability or any other unavoidable circumstances.
Through TeachSpeech, we’re trying to provide a platform for speech therapy that’s easy, repeatable, standardized and ubiquitously available. The Graphical User Interface (GUI) has been designed as simple as possible for the maximum User Experience (UX) since the target audience includes the adults as well as children.

The interactive and simplest Usder Interface (UI) of TeachSpeech

At first the user is asked to read a paragraph through microphone. The user is also provided with the flexibility of editing the predefined paragraph & test the accuracy level of his/her speech. After reading the paragraph, the result page will appear containing the overall accuracy determined by the algorithm, the count of the correctly pronounced words, the count of incorrectly pronounced words. It will also give the user the word wise accuracy percentage through data presentation.
Besides text comparison, it will also provide the time taken to read the paragraph which will determine the rate of uttered words per minute means one’s speech speed.

The results after speech recognition & executing algorithms.

JavaScript, HTML and CSS (Cascading Style Sheet) have been used to build this application. The application also supports 32 languages. The fuzziness algorithm has been used in the comparison between the text that has been produced by using the web speech API and the given text. There are very few constraints such as the first letter of every sentence must be of higher case in order to get the perfect match accuracy and the user has to put a space before starting another sentence while editing the predefined message. The rules are shown in the front page. It was a team project. I had four other team members from Clark University, Massachusetts and Ahsanullah University of Science & Technology. Our instructor was M Ehsan Hoque, who completed his PhD in MIT Media Lab and currently working as an Assistant Professor of Computer Science at University of Rochester.

The project was also featured in StartUp Dhaka Asia. Click here to view the report published on it. Cheers!

Brain storming before the project

Mr Nazmus Saquib testing our project

Mr Nazmus Saquib testing speech recognition accuracy

UI of TeachSpeech

Zara Mahbub, Senior Vice President, Head of Communication & Service Quality at BRAC Bank Limited, testing our speech recognition and stuttering oral test

TeachSpeech on the final demo day