bvp implementation

config.py

@@ -2,7 +2,7 @@ DEBUG = False
 BR_FS = 18
 ACC_FS = 100
 IMU_FS = 120
-PPG_FS = 100
+PPG_FS = 64
 
 ACC_THOLD = 10
 WIN_THOLD = 0.03

modules/digitalsignalprocessing.py

@@ -17,7 +17,7 @@ from skimage.feature import corner_harris, corner_shi_tomasi, peak_local_max
 from ssqueezepy import cwt as sqz_cwt
 
 from config import WIN_THOLD, WINDOW_SIZE, WINDOW_SHIFT, MQA_THOLD, ACC_THOLD
-from config import ACC_FS, IMU_FS, FS_RESAMPLE, BR_FS
+from config import ACC_FS, IMU_FS, FS_RESAMPLE, BR_FS, PPG_FS
 from config import MIN_RESP_RATE, MAX_RESP_RATE
 
 def butter_lowpass(lowcut, fs, order=5):
@@ -223,7 +223,7 @@ def acc_signal_processing(data, fs:int=100):
 
     # ma = accel - movingaverage(accel, 3)
 
-    data_sd = td_scaler(accel, axis=0)
+    data_sd = std_scaler(accel, axis=0)
     # data_sd = attenuate_edges(data_sd, 20)
 
     # mask = np.abs(data_sd) > thold
@@ -258,6 +258,14 @@ def imu_signal_processing(data, fs:int=IMU_FS):
     ma = movingaverage(bp, 8, axis=0)
     return ma
 
+def bvp_signal_processing(data, fs:int=PPG_FS):
+    # TODO
+    bp = butter_bandpass_filter(data,
+                                45/60,
+                                150/60, fs=fs, order=2)
+    ma = movingaverage(bp, 4, axis=0)
+    return ma
+
 def roddiger_sp(data=None, fs:int=IMU_FS):
     ''' Run data through the following steps:
           * Second order taylor estimation

modules/utils.py

@@ -146,7 +146,7 @@ def get_inter_feature_hist(df, lbl_str='br', ntop_features=5, nsbjs=30):
 def model_training(mdl_str, x_train, y_train, marker,
                    validation_data=None, overwrite=False,
                    is_regression=False, project_directory=None,
-                   window_size=50, extra_train=20):
+                   window_size=50, extra_train=20, poly_deg=1):
     directory = join(project_directory, '_'.join([mdl_str, marker]))
 
     if validation_data is not None:
@@ -258,11 +258,9 @@ def model_training(mdl_str, x_train, y_train, marker,
 
         if validation_data is None:
             tuner.search(x_train, y_train, validation_data,
-                         validation_split=0.2,
-                         batch_size=hypermodel.batch_size)
+                         validation_split=0.2)
         else:
-            tuner.search(x_train, y_train, validation_data,
-                         batch_size=hypermodel.batch_size)
+            tuner.search(x_train, y_train, validation_data)
 
         if overwrite or not exists(tuner.best_model_path+'.index'):
             mdl = tuner.load_model(is_training=True)
@@ -272,7 +270,6 @@ def model_training(mdl_str, x_train, y_train, marker,
             history = hypermodel.fit(
                 None, mdl, x_train, y_train,
                 validation_data=validation_data, epochs=extra_train,
-                batch_size=hypermodel.batch_size,
                 callbacks=callbacks,
             )
             tuner.save_weights_to_path()
@@ -296,7 +293,7 @@ def model_training(mdl_str, x_train, y_train, marker,
     elif mdl_str == 'linreg':
         print("---LinearRegression---")
         # n to 2 if full set, 1 if M and R
-        poly = PolynomialFeatures(1)
+        poly = PolynomialFeatures(poly_deg)
         x_train_poly = poly.fit_transform(x_train)
         mdl_cls = LinearRegressionClass(marker=marker, directory=directory)
         if overwrite: