Direction of arrival estimation using deep neural network for hearing aid applications using smartphone

Area of Science:

• Signal Processing
• Machine Learning
• Acoustics

Background:

• Deep neural networks (DNNs) offer significant performance improvements in various applications.
• Accurate direction of arrival (DOA) estimation is crucial for enhancing hearing studies and hearing aid functionality.
• Existing DOA methods may require external components or struggle in noisy environments.

Purpose of the Study:

• To propose a novel DNN-based method for estimating the direction of arrival (DOA) of speech sources.
• To develop a cost-effective, stand-alone DOA estimation platform using readily available smartphones.
• To improve the performance of hearing aids and facilitate hearing research through enhanced DOA estimation.

Main Methods:

• The DOA estimation problem is framed as a classification task.
• Magnitude and phase of speech signals are utilized as feature sets for DNN training.
• A DNN model is trained using real speech and noisy speech data recorded on smartphones in diverse noisy environments with low signal-to-noise ratios (SNRs).

Main Results:

• The DNN-based DOA method was successfully implemented and executed in real-time on an Android smartphone.
• Objective and subjective performance evaluations were conducted in both training and unseen environments.
• The proposed method demonstrated superior performance compared to conventional DOA estimation techniques.

Conclusions:

• The developed DNN-based DOA estimation method provides a cost-effective and high-performance solution for hearing aid applications and hearing studies.
• Real-time implementation on smartphones enables practical, stand-alone DOA estimation without external hardware.
• The method's effectiveness in low SNR and noisy conditions highlights its robustness and potential for real-world applications.