Research Article
Locomotion Strategy Selection for a Hybrid Mobile Robot Using Time of Flight Depth Sensor
Table 2
Effects of different feature extraction levels and filtering to SVM and LDA classification accuracies.
| Number of iterations | 2 | 3 | 4 | 5 | Number of features | 20 | 84 | 340 | 1364 | Extraction time (ms) | 1.1 | 2 | 3.7 | 8.3 | Unfiltered data | | | | | LDA | 70.16% | 92.99% | 94.57% | 92.72% | SVM | 77.65% | 90.86% | 95.27% | 96.32% | Filtered data | | | | | LDA | 83.46% | 94.78% | 96.71% | 95.12% | SVM | 92.78% | 97.02% | 98.13% | 98.46% |
|
|