Whisper model adapted for Persian Speech Emotion Recognition with PCA

Researchers have explored methods to improve Speech Emotion Recognition (SER) for low-resource languages like Persian, focusing on the Whisper model. Their study proposes using Principal Component Analysis (PCA) to reduce the dimensionality of Whisper's frame-level embeddings, which decreases trainable parameters and training latency without sacrificing performance. While PCA-based reduction consistently enhanced emotion recognition accuracy, fine-tuning Whisper on Persian Automatic Speech Recognition (ASR) tasks showed only minor improvements for SER, indicating limited transferability of language adaptation to emotion-specific representations. AI

IMPACT This research offers practical insights for efficiently using large pretrained speech models for emotion recognition in low-resource languages.

RANK_REASON Academic paper detailing a study on model adaptation and dimensionality reduction for a specific task. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• Persian
• principal component analysis
• Whisper

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →