GeHirNet: A Gender-Aware Hierarchical Model for Voice Pathology Classification

2025-08-01

AI Summary

Speech pathology classification faces two key challenges: gender-related acoustic bias and severe class imbalance due to scarcity of rare disorder samples. To address these, we propose a gender-aware hierarchical modeling framework comprising two stages: (1) accurate speaker gender identification and extraction of gender-specific acoustic features; and (2) gender-conditioned disease classification. We further introduce novel multi-scale resampling and time-warping data augmentation strategies to mitigate both bias and imbalance. Our model employs ResNet-50 for Mel-spectrogram analysis and is trained on a unified corpus comprising four public datasets. It achieves 97.63% accuracy and 95.25% Matthews Correlation Coefficient (MCC), outperforming the single-stage baseline by 5.0 percentage points—setting a new state-of-the-art. This advancement significantly enhances the clinical viability of AI-driven speech pathology diagnosis.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Natural Language Processing: Speech
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

AI-based voice analysis shows promise for disease diagnostics, but existing classifiers often fail to accurately identify specific pathologies because of gender-related acoustic variations and the scarcity of data for rare diseases. We propose a novel two-stage framework that first identifies gender-specific pathological patterns using ResNet-50 on Mel spectrograms, then performs gender-conditioned disease classification. We address class imbalance through multi-scale resampling and time warping augmentation. Evaluated on a merged dataset from four public repositories, our two-stage architecture with time warping achieves state-of-the-art performance (97.63% accuracy, 95.25% MCC), with a 5% MCC improvement over single-stage baseline. This work advances voice pathology classification while reducing gender bias through hierarchical modeling of vocal characteristics.

Problem

• Classifying voice pathologies despite gender-related acoustic variations
• Addressing data scarcity for rare diseases in voice pathology detection
• Reducing gender bias in AI-based voice pathology classification

Innovation

• Two-stage gender-aware hierarchical classification model
• ResNet-50 on Mel spectrograms for pattern detection
• Multi-scale resampling and time warping augmentation