File size: 15,845 Bytes
ca8e271 f7b4d24 ca8e271 7be20ea ca8e271 45d17fb 32e7d7d 45d17fb ca8e271 45d17fb 7be20ea 45d17fb ca8e271 45d17fb ca8e271 45d17fb ca8e271 45d17fb ca8e271 7be20ea 45d17fb ca8e271 7be20ea ca8e271 bc14906 7be20ea 45d17fb ca8e271 45d17fb 7be20ea 45d17fb ca8e271 45d17fb 7be20ea bc14906 45d17fb ca8e271 bc14906 ca8e271 bc14906 ca8e271 bc14906 ca8e271 bc14906 ca8e271 32e7d7d ca8e271 32e7d7d ca8e271 33db364 bc14906 ca8e271 bc14906 32e7d7d bc14906 32e7d7d bc14906 32e7d7d ca8e271 d9b6f8e ca8e271 33db364 ca8e271 bc14906 ca8e271 bc14906 ca8e271 bc14906 ca8e271 33db364 ca8e271 d9b6f8e ca8e271 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 | import os
from typing import Tuple
import h5py
import numpy as np
import scipy.io
import scipy.signal as signal
from scipy.signal import iirnotch
from tqdm import tqdm
def sequence_to_seconds(seq_len: int, fs: float) -> float:
"""Converts a sequence length in samples to time in seconds.
Args:
seq_len (int): The number of samples in the sequence.
fs (float): The sampling frequency in Hz.
Returns:
float: The duration of the sequence in seconds.
"""
return seq_len / fs
def random_amplitude_scale(
sig: np.ndarray, scale_range: Tuple[float, float] = (0.9, 1.1)
) -> np.ndarray:
"""Applies random amplitude scaling to the input signal.
Args:
sig (np.ndarray): The input signal array of shape (T, D).
scale_range (Tuple[float, float], optional): The range [min, max] for the scaling factor.
Defaults to (0.9, 1.1).
Returns:
np.ndarray: The scaled signal array.
"""
scale = np.random.uniform(*scale_range)
return sig * scale
def random_time_jitter(sig: np.ndarray, jitter_ratio: float = 0.01) -> np.ndarray:
"""Adds random Gaussian noise (jitter) to the input signal.
Args:
sig (np.ndarray): The input signal array of shape (T, D).
jitter_ratio (float, optional): The ratio to scale the noise relative to
each channel's standard deviation. Defaults to 0.01.
Returns:
np.ndarray: The signal with added jitter.
"""
T, D = sig.shape
std_ch = np.std(sig, axis=0)
noise = np.random.randn(T, D) * (jitter_ratio * std_ch)
return sig + noise
def random_channel_dropout(sig: np.ndarray, dropout_prob: float = 0.05) -> np.ndarray:
"""Randomly zeros out channels in the signal based on a probability.
Args:
sig (np.ndarray): The input signal array of shape (T, D).
dropout_prob (float, optional): Probability of dropping each channel.
Defaults to 0.05.
Returns:
np.ndarray: The signal with dropped channels.
"""
T, D = sig.shape
mask = np.random.rand(D) < dropout_prob
sig[:, mask] = 0.0
return sig
def augment_one_sample(seg: np.ndarray) -> np.ndarray:
"""Applies a sequence of random augmentations to a single signal segment.
Args:
seg (np.ndarray): Single signal segment of shape (window_size, n_ch).
Returns:
np.ndarray: The augmented signal segment.
"""
out = seg.copy()
out = random_amplitude_scale(out, (0.9, 1.1))
out = random_time_jitter(out, 0.01)
out = random_channel_dropout(out, 0.05)
return out
def augment_train_data(
data: np.ndarray, labels: np.ndarray, factor: int = 3
) -> Tuple[np.ndarray, np.ndarray]:
"""Augments the training dataset by creating multiple versions of each sample.
Args:
data (np.ndarray): The input dataset of shape (N, window_size, n_ch).
labels (np.ndarray): The corresponding labels of shape (N,).
factor (int, optional): The number of augmented versions to create for each sample.
Defaults to 3.
Returns:
Tuple[np.ndarray, np.ndarray]: A tuple containing:
- The augmented dataset.
- The augmented labels.
"""
if factor <= 0 or data.shape[0] == 0:
return data, labels
aug_segs = [data]
aug_lbls = [labels]
N = data.shape[0]
for i in tqdm(range(N), desc="Augmenting training data"):
seg = data[i] # [window_size, n_ch]
lab = labels[i]
for _ in range(factor):
aug_segs.append(augment_one_sample(seg)[None, ...])
aug_lbls.append([lab])
new_data = np.concatenate(aug_segs, axis=0)
new_labels = np.concatenate(aug_lbls, axis=0).ravel()
return new_data, new_labels
def resample_to_rate(
emg: np.ndarray,
label: np.ndarray,
rerep: np.ndarray,
original_fs: float,
target_fs: float,
) -> Tuple[np.ndarray, np.ndarray, np.ndarray]:
"""Resamples EMG and aligns sample-wise labels/repetition indices.
EMG is resampled with a polyphase anti-imaging filter. Labels and repetition
indices are mapped with nearest-neighbor sampling so they remain discrete.
Args:
emg (np.ndarray): EMG array of shape (T, D).
label (np.ndarray): Sample-wise labels of shape (T,).
rerep (np.ndarray): Sample-wise repetition indices of shape (T,).
original_fs (float): Original sampling frequency in Hz.
target_fs (float): Target sampling frequency in Hz.
Returns:
Tuple[np.ndarray, np.ndarray, np.ndarray]: Resampled EMG, labels, and
repetition indices.
"""
if target_fs <= 0:
raise ValueError("target_fs must be positive")
if np.isclose(original_fs, target_fs):
return emg, label, rerep
from fractions import Fraction
ratio = Fraction(target_fs / original_fs).limit_denominator(1000)
up, down = ratio.numerator, ratio.denominator
emg_resampled = signal.resample_poly(emg, up, down, axis=0)
new_len = emg_resampled.shape[0]
source_idx = np.rint(
np.arange(new_len, dtype=np.float64) * original_fs / target_fs
).astype(np.int64)
source_idx = np.clip(source_idx, 0, len(label) - 1)
return emg_resampled, label[source_idx], rerep[source_idx]
def notch_filter(
data: np.ndarray, notch_freq: float = 50.0, Q: float = 30.0, fs: float = 200.0
) -> np.ndarray:
"""Applies a notch filter to remove power line interference.
Args:
data (np.ndarray): The input signal array of shape (T, D).
notch_freq (float, optional): The frequency to be removed (e.g., 50Hz or 60Hz).
Defaults to 50.0.
Q (float, optional): The quality factor. Defaults to 30.0.
fs (float, optional): The sampling frequency of the signal. Defaults to 200.0.
Returns:
np.ndarray: The filtered signal array.
"""
b, a = iirnotch(notch_freq, Q, fs)
out = np.zeros_like(data)
for ch in range(data.shape[1]):
out[:, ch] = signal.filtfilt(b, a, data[:, ch])
return out
def bandpass_filter_emg(
emg: np.ndarray,
lowcut: float = 20.0,
highcut: float = 90.0,
fs: float = 200.0,
order: int = 4,
) -> np.ndarray:
"""Applies a Butterworth bandpass filter to the EMG signal.
Args:
emg (np.ndarray): The input signal array of shape (T, D).
lowcut (float, optional): Lower bound of the passband in Hz. Defaults to 20.0.
highcut (float, optional): Upper bound of the passband in Hz. Defaults to 90.0.
fs (float, optional): The sampling frequency of the signal. Defaults to 200.0.
order (int, optional): The order of the filter. Defaults to 4.
Returns:
np.ndarray: The bandpass filtered signal array.
"""
nyq = 0.5 * fs
low = lowcut / nyq
high = highcut / nyq
b, a = signal.butter(order, [low, high], btype="bandpass")
out = np.zeros_like(emg)
for c in range(emg.shape[1]):
out[:, c] = signal.filtfilt(b, a, emg[:, c])
return out
def process_emg_features(
emg: np.ndarray,
label: np.ndarray,
rerep: np.ndarray,
window_size: int = 1024,
stride: int = 512,
majority: bool = False,
) -> Tuple[np.ndarray, np.ndarray, np.ndarray]:
"""Segments raw EMG signals into overlapping windows.
Args:
emg (np.ndarray): Raw EMG data of shape (T, n_ch).
label (np.ndarray): Gesture labels of shape (T,).
rerep (np.ndarray): Repetition indices of shape (T,).
window_size (int, optional): Number of samples per window. Defaults to 1024.
stride (int, optional): Number of samples to shift between windows. Defaults to 512.
Returns:
Tuple[np.ndarray, np.ndarray, np.ndarray]: A tuple containing:
- windowed segments (N, window_size, n_ch).
- labels for each window (N,).
- repetition indices for each window (N,).
"""
segs, lbls, reps = [], [], []
N = len(label)
for start in range(0, N - window_size + 1, stride):
end = start + window_size
win = emg[start:end]
window_labels = label[start:end]
window_reps = rerep[start:end]
if majority:
# Majority voting over gestures
labels_u, labels_c = np.unique(window_labels, return_counts=True)
assigned_label = int(labels_u[np.argmax(labels_c)])
else:
# Pick index 0 for labe
assigned_label = window_labels[0]
assigned_rep = window_reps[0]
segs.append(win)
lbls.append(assigned_label)
reps.append(assigned_rep)
return np.array(segs), np.array(lbls), np.array(reps)
def main():
import argparse
args = argparse.ArgumentParser(description="Process EMG data from DB5.")
args.add_argument("--download_data", action="store_true")
args.add_argument("--data_dir", type=str)
args.add_argument("--save_dir", type=str)
args.add_argument(
"--seq_len", type=int, help="Size of the window in samples for segmentation."
)
args.add_argument(
"--stride",
type=int,
help="Step size between windows in samples for segmentation.",
)
args.add_argument(
"--only-41",
action="store_true",
help="If set, only the first 41 gestures (common between DB5, DB7) will be processed.",
)
args.add_argument(
"--data-augment",
action="store_true",
help="If set, data augmentation will be applied to the training set.",
)
args.add_argument(
"--augment-factor",
type=int,
default=3,
help="Number of augmented versions to create for each training sample.",
)
args.add_argument(
"--resample-2khz",
action="store_true",
help=(
"If set, resample EMG from 200 Hz to 2000 Hz before segmentation. "
"Labels and repetition indices are aligned with nearest-neighbor mapping."
),
)
args.add_argument(
"--majority",
action="store_true",
)
args = args.parse_args()
data_dir = args.data_dir
save_dir = args.save_dir
os.makedirs(save_dir, exist_ok=True)
# download data if requested
if args.download_data:
# https://ninapro.hevs.ch/instructions/DB5.html
len_data = range(1, 11) # 1–10
base_url = "https://ninapro.hevs.ch/files/DB5_Preproc/"
# download and unzip
for i in len_data:
url = f"{base_url}s{i}.zip"
os.system(f"wget -P {data_dir} {url}")
os.system(f"unzip -o {data_dir}/s{i}.zip -d {data_dir}")
os.system(f"rm {data_dir}/s{i}.zip")
print(f"Downloaded and unzipped subject {i}\n{data_dir}/s{i}.zip")
original_fs = 200.0
target_fs = 2000.0 if args.resample_2khz else original_fs
window_size, stride = args.seq_len, args.stride
window_seconds = sequence_to_seconds(window_size, target_fs)
print(
f"Sampling rate: {target_fs:.0f} Hz"
+ (f" (resampled from {original_fs:.0f} Hz)" if args.resample_2khz else "")
)
print(f"Window size: {window_size} samples ({window_seconds:.2f} seconds)")
print(f"{args.majority=}")
train_reps = [1, 3, 4, 6]
val_reps = [2]
test_reps = [5]
all_data = {"train": [], "val": [], "test": []}
all_lbls = {"train": [], "val": [], "test": []}
for subj in sorted(os.listdir(data_dir)):
if subj.startswith("h5"):
# Skip folders created by this script when run multiple times
continue
subj_path = os.path.join(data_dir, subj)
if not os.path.isdir(subj_path):
continue
print(f"Processing subject {subj}...")
for mat in sorted(os.listdir(subj_path)):
if not mat.endswith(".mat"):
continue
dd = scipy.io.loadmat(os.path.join(subj_path, mat))
emg = dd["emg"] # [N,16]
label = dd["restimulus"].ravel().astype(int)
rerep = dd["rerepetition"].ravel().astype(int)
# label shift by exercise
# E1: 12 gestures [1-12]
# E2: 17 gestures [1-17]
# E3: 23 gestures [1-23]
if args.only_41:
# Use only E2 and E3 and remap them to labels 1..17 and 18..40
# while keeping rest as label 0.
if "E1" in mat:
continue
if "E2" in mat:
label = np.where(label != 0, label, 0)
elif "E3" in mat:
label = np.where(label != 0, label + 17, 0)
else:
continue
else:
# Flatten the label space to avoid overlap across exercises
if "E2" in mat:
label = np.where(label != 0, label + 12, 0)
elif "E3" in mat:
label = np.where(label != 0, label + 29, 0)
# Filter at the original acquisition rate. Upsampling afterward does not
# create new frequency content, but provides a 2 kHz sample grid when requested.
emg_filt = bandpass_filter_emg(emg, 20, 90, fs=original_fs)
emg_filt = notch_filter(emg_filt, 50, 30, fs=original_fs)
if args.resample_2khz:
emg_filt, label, rerep = resample_to_rate(
emg_filt, label, rerep, original_fs, target_fs
)
# z-score
channel_std = emg_filt.std(axis=0, ddof=1)
channel_std[channel_std == 0] = 1.0
emg_z = (emg_filt - emg_filt.mean(axis=0)) / channel_std
# segment
segs, lbls, reps = process_emg_features(
emg_z, label, rerep, window_size, stride, args.majority
)
# split by repetition index
for seg, lab, rp in zip(segs, lbls, reps):
if rp in train_reps:
all_data["train"].append(seg)
all_lbls["train"].append(lab)
elif rp in val_reps:
all_data["val"].append(seg)
all_lbls["val"].append(lab)
elif rp in test_reps:
all_data["test"].append(seg)
all_lbls["test"].append(lab)
# stack, augment train, transpose, save, and print stats
stats = {}
for split in ["train", "val", "test"]:
X = np.stack(all_data[split], axis=0) # [N, window_size, ch]
y = np.array(all_lbls[split], dtype=int)
if split == "train" and args.data_augment:
X, y = augment_train_data(X, y, factor=args.augment_factor)
# transpose to [N, ch, window_size]
X = X.transpose(0, 2, 1)
# save
with h5py.File(os.path.join(save_dir, f"{split}.h5"), "w") as hf:
hf.create_dataset("data", data=X)
hf.create_dataset("label", data=y)
# compute stats
uniq, cnt = np.unique(y, return_counts=True)
stats[split] = (X.shape, dict(zip(uniq.tolist(), cnt.tolist())))
# print stats
for split, (shape, dist) in stats.items():
print(f"\n{split} → X={shape} [N, C, T]\nlabel distribution:")
for lab, count in dist.items():
print(f" label {lab}: {count} samples")
if __name__ == "__main__":
main()
|