<![CDATA[Background: Obstructive sleep apnea (OSA) is a prevalent global health issue, yet current monitoring methods are often inaccessible or impractical for routine, at-home use. Objective: This study addresses the need for simple, non-invasive tools for sleep quality assessment. Methods: We developed and evaluated a novel, browser-based snoring detection application leveraging TensorFlow.js. A binary classification model was trained on a balanced dataset of snoring and background noise audio. A key feature of the application is its client-side architecture, where all audio processing and model inference occur locally on the user's device, ensuring real-time performance and preserving user privacy. The model's performance was validated on a holdout test set using standard classification metrics. Results: The model demonstrated robust performance, achieving an overall accuracy of 94.12%, a sensitivity of 95.00%, and a specificity of 93.22% in distinguishing snoring from ambient noise. The application successfully generated useful session statistics for users, including total snoring duration, frequency, and the percentage of snoring time during a monitoring session. Conclusion: This study validates the use of a browser-based AI system as a reliable, scalable, and privacy-preserving tool for sleep quality monitoring. While not a diagnostic instrument, the application serves as a highly accessible preliminary screening and awareness tool. This approach represents a significant step toward democratizing sleep health monitoring and empowering individuals to take an active role in managing their well-being.]]>
