Multi-Stream Audio Pressure Sensor Fusion for Impact Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for classifying external impacts on windows or access openings of enclosed structures, such as those using barometric pressure sensors and microphones, face challenges in distinguishing between different types of events and are prone to false alarms due to their simplicity.
Innovation Solution
A classification system that combines audio and pressure signals using a multi-stream method, where pre-processed audio and pressure signals are converted into feature vectors, combined, and then classified using a machine learning algorithm, specifically an artificial neural network with a softmax output.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a simple threshold-based algorithm is used for classification, then the system is simple and inexpensive, but it cannot distinguish between different types of events and generates too many false alarms
Solution Approach 1:
The patent segments the classification process into multiple independent streams: one for audio signal processing and one for pressure signal processing. Each stream extracts features independently and produces separate classification results, which are then combined through sensor fusion to achieve reliable event detection while maintaining algorithmic simplicity
Solution Approach 2:
The patent merges the outputs of the audio stream and pressure stream through sensor fusion using a logical AND operation. This combination allows the system to distinguish between different event types by requiring consistent signals from both sensors, thereby reducing false alarms while keeping the overall system architecture simple
2Device complexity
If a majority rule is applied to combine pressure and microphone signals, then the system remains simple, but it does not make efficient use of the two available streams and compromises classification performance
Solution Approach 1:
The patent maintains separate processing streams for audio and pressure signals, extracting features independently in each stream. This segmentation allows each sensor type to be optimized for its specific characteristics while preserving the simplicity of the overall fusion mechanism through a straightforward logical AND operation on the final results
Data Source
AI summary
In accordance with an embodiment, a system includes: an audio signal pre-processor; a pressure signal pre-processor; an audio signal feature processor; a pressure signal feature processor; a feature combining processor; and a classification processor configured for classifying the external impact on a window or access opening of an enclosed structure by classifying an audio feature and pressure feature vector in order to produce a classification output; wherein the classification processor is configured for executing a first machine learning algorithm, wherein the audio feature and pressure feature vector is fed to an input layer of the first machine learning algorithm, and wherein the classification output is based on an output of the first machine learning algorithm.