MERaLiON-GR: Speech Gender Recognition Model for English and SEA Languages

2026-08-05

AI Summary

This work addresses cross-lingual speaker gender recognition across English and seven Southeast Asian languages by proposing an efficient multilingual gender classification system based on the pretrained large model MERaLiON-SpeechEncoder-2. The approach leverages parameter-efficient fine-tuning via Low-Rank Adaptation (LoRA) and integrates multi-scale ECAPA-TDNN, attention pooling, and a lightweight linear classifier. Notably, it represents the first successful combination of LoRA with ECAPA-TDNN for multilingual speech gender recognition. The proposed method substantially enhances cross-lingual generalization and achieves state-of-the-art performance on both full utterances and short segments, outperforming existing approaches—including Vox-Profile and large audio language models—across all evaluated settings.

Abstract

We present MERaLiON-GR, a speech gender recognition system that performs binary classification (female / male) on English and Southeast Asian (SEA) languages. The model finetunes MERaLiON-SpeechEncoder-2, a large conformer based transformer pre-trained on a broad speech corpus, and applies parameter efficient fine-tuning via Low-Rank Adaptation (LoRA) to adapt the encoder to the gender recognition task, and appends a multi-scale ECAPA-TDNN down stream network with attention pooling and a lightweight linear classifier. Extensive evaluations across multilingual Singaporean and Southeast Asian languages (English, Chinese, Malay, Tamil, Thai, Vietnamese, Indonesian, and Khmer) show that MERaLiON-GR consistently surpasses the state-of-the-art gender recognition model Vox-Profile and a large Audio-LLM, in both full-utterance and segment level evaluation modes. The results underscore the value of dedicated speech models in achieving accurate paralinguistic understanding and strong cross-lingual generalization.

Problem

• speech gender recognition
• multilingual
• Southeast Asian languages
• paralinguistic analysis
• cross-lingual generalization

Innovation

• Low-Rank Adaptation
• Conformer-based Transformer
• ECAPA-TDNN
• Cross-lingual Generalization
• Parameter-Efficient Fine-tuning