CogSR: Semantic-Aware Speech Super-Resolution via Chain-of-Thought Guided Flow Matching

2025-12-18

AI Summary

Existing generative models for super-resolution reconstruction of historical and surveillance audio at extremely low sampling rates (e.g., <4 kHz) suffer from severe semantic hallucination due to critically insufficient acoustic cues. To address this, we propose a semantics-guided cognitive restoration framework: (1) introducing Chain-of-Thought (CoT) reasoning as an explicit semantic anchor to model linguistic structure; (2) jointly optimizing speech content fidelity and speaker identity consistency via rectified flow-based generation coupled with an acoustic identity constraint module. Our method significantly mitigates ambiguity under severe degradation, achieving a 23.6% improvement in word-level accuracy and a 4.8 dB gain in PSNR for high-frequency details. This work presents the first end-to-end solution for high-fidelity audio restoration that simultaneously preserves linguistic semantics and acoustic authenticity—enabling reliable applications in forensic audio analysis and digital humanities.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Deep Generative Models & Autoencoders
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Large language models for search
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Applying speech super-resolution (SR) to recordings with severely low sampling rates is a critical challenge in digital archiving and investigative audio recovery. In these scenarios, the input lacks essential acoustic cues. Consequently, existing generative models often fail; without sufficient context, they hallucinate phonetic content, guessing words based on probability rather than meaning. To address this, we propose CogSR, a framework designed specifically for high-precision, offline restoration. Our approach shifts the focus from simple signal mapping to cognitive reconstruction. By integrating a Large Audio-Language Model, we employ Chain-of-Thought reasoning to act as a semantic anchor, while explicit acoustic priors ensure the speaker's identity remains consistent. This guides a Rectified Flow backbone to synthesize high-frequency details that are not only realistic but linguistically accurate. Evaluations show that CogSR effectively eliminates ambiguity in severe degradation regimes, making it a robust solution for restoring high-value legacy and surveillance audio.

Problem

• Restores low-sampling-rate speech with semantic accuracy
• Prevents hallucination of phonetic content in degraded audio
• Ensures speaker identity consistency in super-resolution reconstruction

Innovation

• Chain-of-Thought reasoning guides semantic reconstruction
• Rectified Flow backbone synthesizes realistic high-frequency details
• Large Audio-Language Model integrates acoustic and linguistic priors