Audio Anomaly Detection Using PCA Augmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Generating a training data set with sufficient variance for anomaly detection in mechanical devices is challenging, leading to reduced accuracy in identifying device malfunctions, as existing methods often lack the necessary quantity and variability of audio samples to effectively differentiate between normal and abnormal sounds.
Innovation Solution
The approach involves using principal component analysis (PCA) to generate additional audio samples with higher variance by augmenting original audio samples, creating a matrix that includes both original and augmented samples, and selecting principal components based on variance to enhance the detection model's accuracy, thereby reducing false positives and improving anomaly detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If original audio samples are collected from mechanical devices, then the training data set can be generated, but the quantity and variance of samples are insufficient to provide high quality detection
Solution Approach 1:
The patent creates synthetic copies of original audio samples by generating principal components from the original data and reconstructing augmented audio samples. This copying approach multiplies the limited original samples into a larger training dataset, directly addressing the insufficient quantity of audio samples while maintaining the characteristic variance needed for accurate anomaly detection
Solution Approach 2:
The patent applies parameter changes by transforming the original audio samples through principal component analysis, where the data is projected into a new coordinate system defined by principal components. This transformation changes the parameter representation of the audio data, creating augmented samples with varied characteristics that increase both quantity and variance of the training dataset
2Reliability
If principal component analysis is used to generate augmented samples, then the variance of the training data set increases, but the computational complexity increases
Solution Approach 1:
The patent extracts only the essential features from the original audio samples by identifying and retaining the top principal components that capture the majority of variance. This extraction approach reduces computational complexity by focusing on the most significant components while still generating sufficient variance in the augmented samples to maintain high detection accuracy
Solution Approach 2:
The patent generates a controlled number of augmented samples by applying partial action - creating enough augmented data to significantly improve training dataset variance and detection accuracy, but not excessively multiplying the samples to the point of unnecessary computational burden. The method balances the trade-off between variance improvement and computational cost
Data Source
AI summary
Example implementations relate to audio samples to detect device anomalies. For example, computing device, comprising: a processing resource and a non-transitory computer readable medium storing instructions executable by the processing resource to: generate a matrix of audio information for a plurality of audio samples of a device, select audio information from one of the plurality of audio samples, generate a plurality of principal components for the selected audio information utilizing a principal component expansion, select a principal component from the plurality of principal components based on a quantity of variance, and detect an anomaly of the device based on a comparison between a real time audio sample of the device and the selected principal component.