What is AI-Defined Hearing (ADH)?
AI-Defined Hearing (ADH) represents a groundbreaking shift from traditional rule-based signal processing to dynamic, self-learning auditory systems. Building on the foundation of Software-Defined Hearing (SDH), ADH empowers Artificial Intelligence (AI) algorithms to personalize and adapt sound processing in real-time, creating unparalleled hearing experiences.
Key Features of ADH (AI-Defined Hearing)
Machine Learning and Cloud Connectivity: ADH integrates machine learning and cloud connectivity to transform hearing aids into intelligent AIoT platforms that evolve with the user's preferences and environment.
Personalized and Adaptive Experiences: ADH creates highly personalized and adaptive experiences by leveraging AI algorithms to analyze user behavior and adjust sound processing accordingly.
Lyratone's ADH Architecture
Lyratone's ADH architecture extends the SDH framework, utilizing a Cloud-Edge-Device synergy to provide "Hearing as a Service" (HaaS). The architecture consists of four layers:
Cloud Layer (Big Data & AI Learning)
Global Brain: The cloud layer serves as the global brain, analyzing anonymized data from hundreds of thousands of users to train sophisticated hearing models and acoustic prescriptions.
Large Language Models (LLM) and Big Data: The cloud layer utilizes LLMs and big data to create advanced hearing models and acoustic prescriptions that are pushed to devices for continuous performance enhancement.
Internet Application Layer (Control & Logic)
Communication Gateway: The internet application layer acts as the communication gateway, utilizing AI to optimize the interactive fitting process.
Smartphone-Based Self-Testing and Predictive Personalization: The layer enables smartphone-based self-testing, predictive personalization, and smart features like voice command control and real-time multi-language translation.
Embedded Application Layer (Real-Time Edge AI)
Real-Time AI Algorithms: The embedded application layer integrates real-time AI algorithms directly into the device's firmware, utilizing Deep Neural Networks (DNN) and CNN models for advanced tasks like neural speech enhancement, intelligent noise suppression (up to 12dB), and adaptive feedback management.
Hardware Layer (System-on-Chip)
High-Performance SoC Platforms: The hardware layer consists of high-performance SoC platforms (utilizing 6nm to 14nm process nodes), providing the massive computational power required to run complex AI inference models at the "edge" with ultra-low latency (<15ms).