Table 4 Compilation of the feature extraction processes and top-performing classifiers on the US8K_AV dataset.
From: A dataset for environmental sound recognition in embedded systems for autonomous vehicles
Features - Classifier | Experiment 1: Sliding window (1 s) | Experiment 2: Single window (4 s) | ||
|---|---|---|---|---|
Accuracy | (Std) | Accuracy | (Std) | |
Handcrafted - SVM | 78.96% | (1.82%) | 83.14% | (2.21%) |
Handcrafted - LR | 78.87% | (1.49%) | 82.27% | (2.74%) |
Handcrafted - RF | 75.59% | (2.72%) | 79.94% | (2.63%) |
Handcrafted - ANN | 77.93% | (3.29%) | 83.35% | (2.91%) |
Handcrafted - CNN 1D | 76.69% | (3.74%) | 82.40% | (2.72%) |
Aggregated - CNN 2D | 80.03% | (2.71%) | 82.78% | (3.60%) |
