EduSync - Empowering Neurodiversity

EduSync - Empowering Neurodiversity

Personalized AI speech therapy for neurodiverse students. We analyze pronunciation,offer real-time feedback, provide conversational practice to boost confidence, improve communication & social skills.

Created on 15th September 2024

EduSync - Empowering Neurodiversity

EduSync - Empowering Neurodiversity

Personalized AI speech therapy for neurodiverse students. We analyze pronunciation,offer real-time feedback, provide conversational practice to boost confidence, improve communication & social skills.

The problem EduSync - Empowering Neurodiversity solves

About 1 in 12 children experience speech, language, or voice disorders, most commonly between the ages of 3 and 6. By first grade, around 5% of children display noticeable speech disorders such as stuttering and dysarthria, often without a known cause. These challenges can significantly affect communication, confidence, and social interactions. Traditional speech therapy, while effective, is often limited in accessibility, affordability, and the ability to provide personalized, real-time feedback, leaving many children without the support they need.

The Problem Code Blooded is Tackling:
Children with speech disorders frequently struggle with specific phonemes and sounds. Conventional methods may not offer immediate, tailored feedback during practice, and many children lack the confidence to engage in social conversations. Additionally, therapy can be costly and difficult to access, creating barriers to consistent practice and improvement.

How Our Web App Solves the Problem:
Our app uses AI to analyze children's pronunciation during an initial test of letters and words, identifying problematic phonemes. Based on this analysis, it offers personalized practice sessions with real-time feedback to help children improve specific sounds.
Our conversational AI engages children in dialogue, such as asking them to introduce themselves, helping build confidence and social skills while generating reports on their progress.
Additionally, we provide 3D models showing how to form challenging sounds, offering visual guidance that enhances learning. This combination of AI-driven feedback, conversational practice, and visual aids empowers children to practice effectively and improve their speech in an engaging, accessible environment.

Challenges we ran into

Building our application with a diverse tech stack came with several key challenges. Here's how we addressed them:

Gen AI Prompting and Text to Speech Integration :
Configuring the Gen AI model for voice-based tasks and crafting effective prompts was a significant challenge. We needed to ensure the model could process voice inputs and generate accurate responses while aligning prompts with user queries. Report generation added complexity, requiring us to integrate data and ensure output accuracy. Through careful prompt engineering and optimized data workflows, we resolved these challenges.

PyAudio and OpenAI Whisper Integration :
Integrating PyAudio with the Whisper model for speech-to-text conversion was tough, particularly in handling text variations. We didn't fine-tune Whisper but instead focused on backend optimization to compare text accurately. Implementing natural language processing techniques helped improve accuracy in text analysis, ensuring robust performance.

Scalability and Video Asset Management :
Large Blender video files were too big to store on GitHub. To address this, we hosted them on AWS S3 and embedded the links in our documentation. This solution allowed us to manage large assets without performance issues while offering greater flexibility for video manipulation.

Overall, overcoming these technical challenges strengthened our application and demonstrated our ability to deliver a high-quality, scalable solution.

Tracks Applied (1)

GaiaNet

GaiaNet is a decentralized computing infrastructure that empowers individuals and businesses to create, deploy, scale, a...Read More

Hackquest

Discussion

Builders also viewed

See more projects on Devfolio