Auto-encoder Speech Feature Extraction for Infant ASD Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for diagnosing autism spectrum disorder in infants and young children face challenges such as subjectivity, time constraints, and complexity in data analysis, with a lack of organized data and low accessibility to diagnosis, making early detection difficult.
Innovation Solution
A deep learning-based device and method that uses an auto-encoder to extract features from speech data, segmenting and classifying autism spectrum disorder by converting input parameters into latent representations and reconstructing speech features for improved classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If diagnostic instruments are used to distinguish children with ASD from TD children, then classification accuracy is improved, but the procedure stability deteriorates due to time constraints and clinician subjectivity
Solution Approach 1:
The system uses automatic speech analysis where the computational model independently processes speech data and generates diagnostic predictions without requiring continuous clinician intervention, thereby maintaining consistent procedure execution while preserving high classification accuracy
Solution Approach 2:
The patent replaces the manual mechanical diagnostic process with an automated deep learning system that processes speech features algorithmically, eliminating human subjectivity and time variability while maintaining or improving classification precision
2Reliability
If acoustic features and classification algorithms are used to detect ASD abnormalities, then detection effectiveness is improved, but analysis complexity increases
Solution Approach 1:
The system extracts and focuses on specific speech acoustic features relevant to ASD detection using auto-encoders, separating the essential diagnostic features from the overall complex speech data, thereby maintaining detection effectiveness while reducing analysis complexity
Solution Approach 2:
The patent transforms raw speech data into compressed latent representations through auto-encoder encoding, changing the parameter space from high-dimensional raw features to lower-dimensional meaningful features, which reduces analysis complexity while preserving detection effectiveness
3Measurement precision
If deep learning models are used to distinguish ASD from TD children, then detection performance is improved, but data accessibility and organization requirements increase
Solution Approach 1:
The system performs preliminary organization and preprocessing of speech data into standardized formats before analysis, and pre-trains auto-encoder models on speech feature extraction, so that when actual diagnostic data is input, the complex deep learning model can operate efficiently without requiring complex data organization at the point of use
Data Source
AI summary
The present invention relates to disorder spectrum diagnosis technology, and more particularly, to a device and method for predicting autism spectrum disorder in infants and young children on the basis of deep learning by using auto-encoder feature representation, wherein autism spectrum disorder can be identified from the speech of infants and young children by using auto-encoder feature representation.