Tested on | NN trained on | SLR calibrated on |
|---|
GMC | ING | GMC | ING |
|---|
GMC | Experiment 1 | Experiment 3 | Experiment 5 | Experiment 7 |
“NN GMC-GMC” | “NN ING-GMC” | “SLR GMC-GMC” | “SLR ING-GMC” |
ING | Experiment 2 | Experiment 4 | Experiment 6 | Experiment 8 |
“NN GMC-ING” | “NN ING-ING” | “SLR GMC-ING” | “SLR ING-ING” |
- Two different ABR threshold finding methods were tested on two different data sets (GMC and ING). The first two columns contain experiments with the two-stage neural network (NN), the last two columns contain experiments with the sound level regression method (SLR). Sub-columns specify the data set that was used for training (NN) or calibration (SLR), respectively. The two rows indicate the data set that was used for testing of the trained NN or the calibrated SLR model. Cells provide the experiment number and the name of the experiment as used in the text. Experiments that use data from the same data set for training/calibration and testing are highlighted in grey
