Automated machine learning based speech classification for hearing aid applications and its real-time implementation...

Area of Science:

• Audio signal processing
• Machine learning applications
• Hearing aid technology

Background:

• Deep neural networks (DNNs) have advanced audio processing, particularly for hearing aids.
• Automated machine learning (AutoML) simplifies DNN model optimization for easier implementation.
• Existing speech processing algorithms can be enhanced for improved speech perception.

Purpose of the Study:

• To develop an AutoML-based voice activity detector (VAD) for real-time smartphone application.
• To improve speech processing in hearing aid devices through enhanced VAD.
• To demonstrate the practical application of AutoML in hearing aid technology.

Main Methods:

• Utilized an AutoML platform to generate a computationally fast classification model.
• Implemented the AutoML-generated VAD model on a smartphone for real-time operation.
• Detailed the steps for real-time implementation on a mobile device.

Main Results:

• The AutoML-based VAD achieved computationally fast performance with minimal processing delay.
• The real-time smartphone application demonstrated practical usability in various noisy environments.
• Experimental results showed improvements over state-of-the-art techniques.

Conclusions:

• AutoML is a significant platform for developing efficient hearing aid applications.
• The developed smartphone VAD app offers practical usability and enhances speech processing.
• This work highlights the successful realization of an AutoML model on a smartphone for hearing aid applications.