CNN Audio Glitch Detection in Information Handling Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Information handling systems often fail to detect audio glitches due to limitations in human testing, such as background noise and fatigue, making it difficult to accurately identify and diagnose audio errors.
Innovation Solution
A system utilizing a convolution neural network (CNN) that receives audio files with and without glitches, trains on user responses to identify patterns, and convolves audio output from information handling systems with a filter to determine if audio glitches are present, using pooling operations and weight/bias determination via gradient descent processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If human testing is used to detect audio glitches, then the testing process is simple and inexpensive, but the detection accuracy is low due to background noise and fatigue
Solution Approach 1:
The patent replaces the mechanical human testing system with an automated electronic testing system using convolutional neural networks. The CNN model processes audio signals digitally, eliminating human limitations such as fatigue and background noise sensitivity, thereby significantly improving detection accuracy while maintaining manageable system complexity through software-based automation.
2Measurement precision
If automated testing systems are implemented, then detection accuracy improves, but the device complexity increases
Solution Approach 1:
The testing system performs self-training by automatically learning from labeled audio data containing both glitch-free and glitch-containing samples. The convolutional neural network adapts to specific audio characteristics through self-supervised learning, reducing the need for complex manual configuration and expert intervention, thereby managing system complexity while achieving high detection accuracy.
Solution Approach 2:
The system performs preliminary training with labeled audio data before actual testing, pre-configuring the neural network to recognize glitch patterns. This preliminary action prepares the system in advance, reducing the complexity of real-time decision-making during actual audio testing operations.
3Productivity
If multiple audio files are tested manually, then comprehensive coverage is achieved, but the testing time increases due to fatigue and background noise
Solution Approach 1:
The patent replaces manual human testing with automated electronic testing using convolutional neural networks. The system can process multiple audio files simultaneously without fatigue or background noise interference, dramatically improving testing efficiency and reducing the time required to achieve comprehensive audio coverage.
Solution Approach 2:
The automated testing system operates continuously without interruption, processing audio files in sequence or parallel without the breaks and fatigue that limit human testers. This continuous operation maximizes productivity while minimizing total testing time required to evaluate all audio outputs.
Data Source
AI summary
In one or more embodiments, one or more system, methods, and/or processes may determine, based at least on the user responses from users that listen to audio files, first portions of the audio files that include at least one audio glitch and second portions of the audio files that do not include the at least one audio glitch; may determine values of a filter, of a convolution neural network (CNN), based at least on the first portions and the second portions of the audio files; may provide audio produced by an information handling system (IHS) to the CNN; may determine, based at least on data from convolving the audio produced by the IHS with the filter and output data from the CNN, if the IHS has produced an audio glitch; and may provide information indicating whether or not the IHS has produced the audio glitch.