Advanced Noise Reduction Using Canonical Correlation Deep Learning (CCDL) for Real-Time Audio Enhancement
M Amareswar, Nishitha Digadari, Bhagwan Sahu, Aakiti Manikanta Reddy · Journal of Science Engineering Technology and Management Sciences · 2025
Noise reduction plays a vital role in enhancing audio clarity across telecommunications and broadcasting systems, where maintaining speech intelligibility under diverse noise conditions is essential.Traditional approaches like spectral subtraction rely on estimating and removing stationary noise components from the signal spectrum; while effective in stable environments, they often introduce audible artifacts and perform poorly in dynamic or real-time scenarios involving non-stationary noise.To address these challenges, this work proposes a novel Canonical Correlation Deep Learning (CCDL) method for advanced noise reduction that dynamically adapts to varying acoustic environments.The CCDL approach leverages deep learning capabilities alongside canonical correlation analysis to extract meaningful relationships between noisy and clean signals, enabling intelligent parameter adjustment in real time.Its Canonical Correlation Bandpass (CCB) configuration focuses on enhancing critical speech frequency bands while suppressing irrelevant noise, thereby preserving the naturalness and integrity of the audio.Designed for real-time adaptability, CCDL minimizes latency and artifacts, making it highly suitable for applications such as mobile communications, live broadcasts, and voice-based interfaces.Comprehensive evaluations using performance metrics such as Signal-to-Noise Ratio (SNR), Perceptual Evaluation of Speech Quality (PESQ), and Log-Spectral Distance (LSD) demonstrate the proposed method's superiority over conventional techniques, highlighting its effectiveness in delivering clearer, more intelligible audio.CCDL marks a significant advancement in adaptive noise reduction, offering a robust, real-time solution to overcome the limitations of traditional systems in modern communication networks.