A robust DCT-SVD-based framework for wide band speech reconstruction over legacy telephone networks
Аннотация
Wideband (WB) speech, spanning 0–8 kHz, is not transmitted over Public Switch Telephone Networks (PSTNs) due to spectrum limitations. Instead, only Restricted Band (RB) speech up to 4 kHz is supported, which severely degrades speech quality, particularly for older people. This work introduces a novel hybrid steganographic method that combines Discrete Cosine Transform (DCT) and Singular Value Decomposition (SVD) to improve the quality of RB speech while maintaining compatibility with current PSTNs. The method embeds Spread Code Excited Linear Prediction (CELP) parameters of the Enhanced Band (EB) portion of speech (4–8 kHz) into RB speech without compromising clarity, and later reconstructs a high-quality Reconstructed Wideband (RWB) signal at the receiver. Objective evaluations confirm the effectiveness of the approach: a WB Perceptual Evaluation of Speech Quality (WB-PESQ) score of 4.12 demonstrates superior WB speech quality, while an RB-PESQ score of 4.19 highlights excellent perceptual transparency compared to traditional methods. Moreover, the technique maintains Mean Square Error (MSE) values below $$4.98*{10^{ - 8}}$$ across Signal to Noise Ratio (SNR) levels from 15 to 35 dB, confirming resilience to channel and quantization noise. Subjective listening tests further validate the improvements, particularly for elderly users.