Authors: João Victor Carlet¹, Ana Clara Ricci Benetti Boldrin¹, Vagner Guilherme Figueira Neto¹, João Navarro Soares¹, Deborah Viviani Ferrari¹, Ben-Hur Viana Borges¹
¹University of São Paulo
Background: Hearing loss affects 1.5 billion people and is associated with adverse outcomes in health, cognition, and quality of life). In Brazil, 26% of the population has some degree of hearing loss, and 14 million require hearing aid. However, hearing aids are underutilized due to high costs, limited funding, and proprietary technology. Open research initiatives have emerged offering accessible platforms, but often rely on bulky, power-intensive processors and lack user-friendly clinical tools. Engineers typically excel in embedded development but lack clinical infrastructure, while audiologists hold diagnostic expertise but limited programming skills. This technical-clinical gap hinders innovation.
Method: We propose Tiresias, an open-source hearing aid development platform based on consumer-grade tools. It integrates a Nordic processor and an Analog Devices audio processor, combining Bluetooth connectivity with low-power, dedicated audio processing. Hearing functions—like filtering and compression—are created using the Analog Devices visual programming suite, while control logic is implemented in C using Nordic’s. This design shifts signal processing to the audio processor, lowering power usage and development complexity.
Results: A battery-powered circuit board was designed to support multichannel audio processing. Firmware enables dynamic audio configurations. Core algorithms were implemented, and filtering and compression were preliminarily tested, confirming the platform’s feasibility for rapid, wearable hearing aid development.
Conclusion: Tiresias meets the constraints for wearable use while supporting fast prototyping. Next steps include acoustic characterization, enclosure design, and clinical validation. A custom desktop tool will provide audiological tuning, over-the-air updates, and binaural synchronization. A neural network–based sound scene classifier will be added for context-aware processing.
