PhoenixCodec: Taming Neural Speech Coding for Extreme Low-Resource Scenarios

2025-10-24

AI Summary

Addressing the challenge of balancing efficiency and quality in neural speech coding under extreme low-resource constraints (<700 MFLOPs, <30 ms latency, dual-bitrate support at 1/6 kbps), this paper proposes an efficient neural speech codec framework. Methodologically, it introduces: (1) an optimized asymmetric frequency-time encoder-decoder architecture that alleviates resource dispersion bottlenecks in conventional decoders; (2) a cyclic calibration and refinement (CCR) training strategy to enhance waveform reconstruction fidelity; and (3) noise-invariant fine-tuning to improve robustness in challenging acoustic conditions (e.g., real-world noise and reverberation). Evaluated on the LRAC 2025 Challenge, the framework achieves third place overall and attains state-of-the-art intelligibility performance—both on clean speech and under realistic noisy/reverberant conditions—at 1 kbps.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Natural Language Processing: Speech
• Machine Learning: Learning on the Edge & Model Compression

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

This paper presents PhoenixCodec, a comprehensive neural speech coding and decoding framework designed for extremely low-resource conditions. The proposed system integrates an optimized asymmetric frequency-time architecture, a Cyclical Calibration and Refinement (CCR) training strategy, and a noise-invariant fine-tuning procedure. Under stringent constraints - computation below 700 MFLOPs, latency less than 30 ms, and dual-rate support at 1 kbps and 6 kbps - existing methods face a trade-off between efficiency and quality. PhoenixCodec addresses these challenges by alleviating the resource scattering of conventional decoders, employing CCR to escape local optima, and enhancing robustness through noisy-sample fine-tuning. In the LRAC 2025 Challenge Track 1, the proposed system ranked third overall and demonstrated the best performance at 1 kbps in both real-world noise and reverberation and intelligibility in clean tests, confirming its effectiveness.

Problem

• Optimizing neural speech coding for extreme low-resource conditions
• Balancing efficiency and quality under strict computational constraints
• Enhancing robustness in noisy environments and intelligibility tests

Innovation

• Optimized asymmetric frequency-time architecture integration
• Cyclical Calibration and Refinement training strategy implementation
• Noise-invariant fine-tuning for enhanced robustness