{ "sample_rate_hz": 20000, "capture_window_ms": 1600, "n_mfcc": 20, "n_mels": 40, "n_fft": 512, "hop_length": 256, "fmin": 125, "fmax": 7500, "n_segments": 8, "channels": [ "mfcc", "delta", "delta2" ], "global_stats": [ "mean", "std", "p5", "p95" ], "segment_stats": [ "mean", "std" ], "full_dim": 1200, "clip_sigma": 4.0, "note": "1200 = 60 streams (20 mfcc + 20 delta + 20 delta2) x (4 global + 8*2 segment) stats", "input_dim": 48, "select_k": 48, "hidden": 32, "num_classes": 18, "labels": [ "_silence_", "_unknown_", "power_on", "power_off", "speed_1", "speed_2", "speed_3", "speed_4", "speed_5", "speed_6", "mode_boost", "mode_nature", "mode_reverse", "mode_smart", "timer_off", "timer_2h", "timer_4h", "timer_8h" ], "val_float_acc": 0.9346, "val_int8_acc": 0.9308, "seed": 1, "int8_bytes": 4400 }