Speaker Authentication Using Neuroevolution Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting time series patterns, such as audio patterns, often rely on feature extraction which can reduce input dimensionality but may also lead to misidentifications by not capturing specific characteristics of speakers or events, especially in speaker authentication tasks.
Innovation Solution
The method employs neuroevolution of augmenting topologies (NEAT) to build specific artificial neural networks for each time series pattern, allowing them to extract relevant features directly from raw input signals and minimizing topology for efficient computing resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If feature extraction is used to reduce input dimensionality, then processing efficiency is improved, but detection accuracy deteriorates due to loss of specific speaker characteristics
Solution Approach 1:
The patent extracts only the essential temporal patterns and characteristics from raw audio signals that are necessary for speaker identification, rather than extracting comprehensive features. This selective extraction maintains critical speaker-specific information while reducing overall data dimensionality, thus balancing processing efficiency with detection accuracy.
Solution Approach 2:
The patent introduces an intermediate processing layer that works directly with raw audio signals to identify temporal patterns before final speaker classification. This intermediary approach avoids the need for traditional feature extraction while maintaining processing efficiency through pattern-based rather than feature-based processing.
2Device complexity
If traditional feature extraction methods are used, then processing complexity is reduced, but speaker-specific characteristics are lost leading to misidentifications
Solution Approach 1:
The patent replaces traditional mechanical feature extraction systems with a neural network-based pattern recognition system. The neural network automatically learns and extracts relevant temporal patterns directly from raw audio signals, eliminating the need for manual feature engineering while preserving speaker-specific characteristics and improving identification reliability.
3Measurement precision
If neuroevolution of augmenting topologies is used to build specific neural networks for each pattern, then detection accuracy is improved, but computational resources increase
Solution Approach 1:
The patent segments the speaker identification task into multiple specialized neural networks, each trained to detect specific temporal patterns in audio signals. This segmentation allows each network to be optimized for particular pattern types, improving overall detection accuracy while enabling efficient resource allocation by activating only relevant networks for each input signal.
Solution Approach 2:
The patent employs dynamic neural network topologies that are adapted and optimized during training through neuroevolution. The networks dynamically adjust their structure and parameters to efficiently capture temporal patterns, achieving high detection accuracy while minimizing computational overhead through optimized network architectures.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
According to a first aspect of the present disclosure, a method for facilitating the detection of one or more time series patterns is conceived, comprising building one or more artificial neural networks, wherein, for at least one time series pattern to be detected, a specific one of said artificial neural networks is built. According to a second aspect of the present disclosure, a corresponding computer program is provided. According to a third aspect of the present disclosure, a non-transitory computer-readable medium is provided that comprises a computer program of the kind set forth. According to a fourth aspect of the present disclosure, a corresponding system for facilitating the detection of one or more time series patterns is provided.