Development of a software that assists speech therapy in Urdu
About 13% of the country?s population suffers from speech and communication disorders. Estimated stuttering population in Pakistan is close to 100 per 10,000 populations. Along with communication disability people cannot have access to speech therapy clinic
2025-06-28 16:32:03 - Adil Khan
Development of a software that assists speech therapy in Urdu
Project Area of Specialization Artificial IntelligenceProject SummaryAbout 13% of the country’s population suffers from speech and communication disorders. Estimated stuttering population in Pakistan is close to 100 per 10,000 populations. Along with communication disability people cannot have access to speech therapy clinics. Even if they do there are not many speech therapists available to patients and charges are usually high ranges between Rs. 3,000 to Rs. 5,000 per visit, and the patient has to make regular weekly visits. A lot of Softwares and mobile applications have been made for other languages but not in Urdu language. Keeping in view the problem we aim to design a computer based speech therapy that would help people to evaluate their speech defects starting form word level to paragraphs. For that we have gathered Urdu words according to the frequency distribution of urdu sounds most frequently used. An Algorithm has been designed with the help of Machine learning that would check/evaluate the defected voice signals and allow the user to do a therapy by his or her own pace. This would make speech therapy in Urdu convenient and accessible for all.
Project ObjectivesWe aim to develop a technique to distinguish between different Urdu sounds. A computer based speech therapy would be developed , aided with an algorithm which would help to test and evaluate the difference between the speech of normal person and a person with speech disability (specifically those who stutter)
And by the end of this project our software would have following attributes;
- User Friendly
- Could assist patients with speech therapy
- Different level with progressive difficulty
- Could be used in Home
- Eliminates requirement to visits hospital
- Improvement on your own pace
Machine Learning would be used to implement the project. Initially the protocol was designed for the data collection. The protocol consisted of a set of words that the subject has to say into a microphone. These words were specific as the frequency distribution of the sounds in these words was according to the frequency distribution of these sounds in normal everyday speeches.
The subjects would have to say these 62 words into the microphone. Once enough data has been collected, machine learning can be used to identify different features of speech. These features would be able to distinguish normal speech from one with stutter. Once trained enough the software would be able to diagnose whether the subject has made improvement through therapy over time or not.
One of these features would be the Short Term Fourier Transform, which takes time bins of a signal and plots the frequencies of signal present in each bin. With the help of other such features, a system can be trained to help with speech therapy.
Benefits of the ProjectAs Urdu is our national language so this project would be user friendly and easy to understand. It would be accessible to everyone and could assist user with speech therapy. Speech therapy sessions are expensive and the person has to regular visit the therapist for appointments, such a software could eliminate this expense and make it convenient for the people to get therapy. User would be able to improve according to his/her own pace with different progressive levels.
Technical Details of Final DeliverableThe final result would be software that uses Short Term Fourier Transform and other features to distinguish normal and defected speech. It will be user-friendly and convenient to use, guiding the user through speech therapy. It would eventually be able to process phrases and sentences to provide better results in diagnosis.
Final Deliverable of the Project Software SystemType of Industry Health Technologies Artificial Intelligence(AI)Sustainable Development Goals Good Health and Well-Being for PeopleRequired Resources| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 22838 | |||
| Streopairs with Headsets | Equipment | 1 | 1098 | 1098 |
| Multimedia EarpphoneMic | Equipment | 5 | 2500 | 12500 |
| Word Cards Color printouts | Miscellaneous | 62 | 20 | 1240 |
| Travelling To and From Hospital for Data Collection | Miscellaneous | 4 | 2000 | 8000 |