Hierarchical Speech Analysis Method for Age, Gender, and Emotion Detection

USPTO published patent application US20260100196A1 assigned to Tencent America LLC, covering a two-stage machine learning method for analyzing speech signals. The first learning stage receives preprocessed speech to detect initial age and gender, while the second learning stage refines these attributes and determines speaker emotion. The system uses hierarchical processing to improve detection accuracy across multiple speaker characteristics.

Technology companies developing speech recognition, emotion detection, or speaker analysis systems should review this patent application for potential implications on future product development and intellectual property strategy. The patent's scope covers AI-based audio processing with applications in customer service, security, healthcare, and consumer electronics.