Supplemental material for the paper "Evaluating Optopalatography Sensor Positions for Silent Command Word Recognition". Authors: Arne-Lukas Fietkau, João Menezes, Peter Birkholz. Accepted for publication at the Elektronische Sprachsignalverarbeitung (ESSV) Conference 2025. Below are the hyperparameter sets which yielded the highest recognition accuracies for each evaluation type and speaker combination. The search space for each hyperparameter is described in the paper. If there are any questions or suggestions, please contact the authors. Evaluation type Speaker Sensors Batch size BLSTM hidden neurons BLSTM dropout rate Single speaker SPK1 All 25 163 0.7 Single speaker SPK1 No lip 20 159 0.3 Single speaker SPK1 No lat. 5 188 0.5 Single speaker SPK1 No lip/lat. 10 192 0.3 Single speaker SPK2 All 25 113 0.3 Single speaker SPK2 No lip 20 171 0.6 Single speaker SPK2 No lat. 20 150 0.6 Single speaker SPK2 No lip/lat. 10 192 0.3 Single speaker SPK3 All 5 188 0.5 Single speaker SPK3 No lip 5 179 0.6 Single speaker SPK3 No lat. 5 246 0.4 Single speaker SPK3 No lip/lat. 5 246 0.4 Single speaker SPK4 All 10 192 0.3 Single speaker SPK4 No lip 50 239 0.8 Single speaker SPK4 No lat. 5 113 0.4 Single speaker SPK4 No lip/lat. 5 246 0.4 Leave-one-spk-out SPK1 All 5 164 0.9 Leave-one-spk-out SPK1 No lip 5 246 0.4 Leave-one-spk-out SPK1 No lat. 10 204 0.6 Leave-one-spk-out SPK1 No lip/lat. 20 243 0.7 Leave-one-spk-out SPK2 All 5 246 0.4 Leave-one-spk-out SPK2 No lip 20 159 0.3 Leave-one-spk-out SPK2 No lat. 5 164 0.9 Leave-one-spk-out SPK2 No lip/lat. 5 246 0.4 Leave-one-spk-out SPK3 All 5 179 0.6 Leave-one-spk-out SPK3 No lip 5 188 0.5 Leave-one-spk-out SPK3 No lat. 5 179 0.6 Leave-one-spk-out SPK3 No lip/lat. 10 204 0.6 Leave-one-spk-out SPK4 All 5 164 0.9 Leave-one-spk-out SPK4 No lip 5 179 0.6 Leave-one-spk-out SPK4 No lat. 10 192 0.3 Leave-one-spk-out SPK4 No lip/lat. 10 121 0.3 Multi-speaker All All 10 218 0.9 Multi-speaker All No lip 10 121 0.3 Multi-speaker All No lat. 20 153 0.6 Multi-speaker All No lip/lat. 20 250 0.3