HASCom: A Heterogeneous Affective-Semantic Communication Framework for Speech Transmission.

Area of Science:

• Wireless Communication
• Speech Processing
• Affective Computing

Background:

• Next-generation wireless networks demand richer communication beyond text.
• Current speech frameworks prioritize semantic accuracy, ignoring crucial affective cues like emotion and tone.
• Natural human interaction relies heavily on conveying and understanding affective information.

Purpose of the Study:

• To propose the Heterogeneous Affective Speech Semantic Communication (HASCom) framework for robust speech transmission over wireless channels.
• To enable the accurate transmission of both linguistic content and affective (emotional) information in speech.
• To improve the naturalness and expressiveness of speech communication in digital systems.

Main Methods:

• Designed a heterogeneous dual-stream architecture separating phoneme-level semantics and emotional embeddings.
• Employed Low-Density Parity-Check (LDPC) coding for reliable discrete semantic transmission.
• Utilized Deep Joint Source-Channel Coding (JSCC) for continuous emotional feature transmission.
• Developed a prior-guided diffusion reconstruction module for enhanced speech recovery at the receiver.

Main Results:

• HASCom significantly outperforms state-of-the-art baselines in speech semantic communication.
• Achieved superior objective semantic similarity and subjective Mean Opinion Score (MOS) at low Signal-to-Noise Ratios (SNRs).
• JSCC modules demonstrated ultra-low inference latency (<0.1 ms), confirming high efficiency.

Conclusions:

• The HASCom framework offers a robust and efficient solution for transmitting expressive speech, including affective information.
• HASCom enhances human-centric communication in wireless systems by preserving emotional nuances.
• The proposed methods show significant potential for practical deployment in advanced communication systems.