Audio Event Detection Using Window-Based Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sound event detection techniques face challenges in accurately identifying audio events in various acoustic environments with different background noise, leading to errors such as false negatives and false positives.
Innovation Solution
A computing system that divides a run-time audio signal into overlapping segments, processes them in the time domain, and feeds the normalized representations to a trained neural network to generate predicted classification scores and probabilities. The system applies smoothing functions and prediction thresholds to improve detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sound event detection is performed using traditional techniques, then the system is simpler to implement, but the detection accuracy and robustness deteriorate in noisy environments
Solution Approach 1:
The audio signal is divided into multiple overlapping segments, where each segment is processed independently by the neural network. This segmentation allows the system to focus computational resources on local patterns while maintaining temporal context through overlaps, improving detection accuracy without requiring excessive complexity in the overall system architecture.
Solution Approach 2:
The system dynamically adjusts the analysis by applying smoothing functions with different window sizes (e.g., 3-segment, 5-segment, 7-segment windows) to the predicted classification scores. This dynamic smoothing adapts to varying noise conditions and event durations, enhancing robustness while keeping the base neural network structure relatively simple.
2Reliability
If overlapping segments are used with smoothing functions, then the robustness and accuracy improve, but the processing time and computational complexity increase
Solution Approach 1:
The system applies smoothing functions with multiple candidate window sizes (3, 5, 7 segments) to ensure sufficient temporal context is captured. By using partial overlaps and selective smoothing rather than processing the entire signal uniformly, the system achieves robust detection without excessive processing time, balancing accuracy and efficiency.
3Measurement precision
If multiple candidate window sizes are evaluated with smoothing functions, then the prediction accuracy improves, but the computational load increases
Solution Approach 1:
The evaluation of multiple candidate window sizes is applied selectively to segments where events are detected, rather than uniformly across the entire audio signal. This segmented approach allows the system to achieve high classification accuracy through multi-scale analysis while reducing overall computational energy consumption by limiting intensive processing to relevant regions.
Data Source
AI summary
A computing system for a plurality of classes of audio events is provided, including one or more processors configured to divide a run-time audio signal into a plurality of segments and process each segment of the run-time audio signal in a time domain to generate a normalized time domain representation of each segment. The processor is further configured to feed the normalized time domain representation of each segment to an input layer of a trained neural network. The processor is further configured to generate, by the neural network, a plurality of predicted classification scores and associated probabilities for each class of audio event contained in each segment of the run-time input audio signal. In post-processing, the processor is further configured to generate smoothed predicted classification scores, associated smoothed probabilities, and class window confidence values for each class for each of a plurality of candidate window sizes.