paradigma 0.3.2__py3-none-any.whl → 0.4.0__py3-none-any.whl

paradigma/classification.py

@@ -0,0 +1,115 @@
+import numpy as np
+import pickle
+
+from pathlib import Path
+from sklearn.base import BaseEstimator
+from typing import Any, Optional
+
+class ClassifierPackage:
+    def __init__(self, classifier: Optional[BaseEstimator] = None, 
+                 threshold: Optional[float] = None, 
+                 scaler: Optional[Any] = None):
+        """
+        Initialize the ClassifierPackage with a classifier, threshold, and scaler.
+
+        Parameters
+        ----------
+        classifier
+            The trained classifier.
+        threshold : float
+            The classification threshold.
+        scaler
+            The trained scaler (e.g., StandardScaler or MinMaxScaler).
+        """
+        self.classifier = classifier
+        self.threshold = threshold
+        self.scaler = scaler
+
+    def transform_features(self, X) -> np.ndarray:
+        """
+        Transform the input features using the scaler.
+
+        Parameters
+        ----------
+        X : np.ndarray
+            The input features.
+
+        Return
+        ------
+        np.ndarray
+            The transformed features.
+        """
+        if not self.scaler:
+            return X
+        return self.scaler.transform(X)
+
+    def predict_proba(self, X) -> float:
+        """
+        Make predictions using the classifier and apply the threshold.
+
+        Parameters
+        ----------
+        X : np.ndarray
+            The input features.
+
+        Return
+        ------
+        float
+            The predicted probability.
+
+        """
+        if not self.classifier:
+            raise ValueError("Classifier is not loaded.")
+        return self.classifier.predict_proba(X)[:, 1]
+    
+    def predict(self, X) -> int:
+        """
+        Make predictions using the classifier and apply the threshold.
+
+        Parameters
+        ----------
+        X : np.ndarray
+            The input features.
+
+        Return
+        ------
+        int
+            The predicted class.
+
+        """
+        if not self.classifier:
+            raise ValueError("Classifier is not loaded.")
+        return int(self.predict_proba(X) >= self.threshold)
+    
+    def save(self, filepath: str | Path) -> None:
+        """
+        Save the ClassifierPackage to a file.
+
+        Parameters
+        ----------
+        filepath : str
+            The path to the file.
+        """
+        with open(filepath, 'wb') as f:
+            pickle.dump(self, f)
+
+    @classmethod
+    def load(cls, filepath: str | Path):
+        """
+        Load a ClassifierPackage from a file.
+
+        Parameters
+        ----------
+        filepath : str
+            The path to the file.
+
+        Return
+        ------
+        ClassifierPackage
+            The loaded classifier package.
+        """
+        try:
+            with open(filepath, 'rb') as f:
+                return pickle.load(f)
+        except Exception as e:
+            raise ValueError(f"Failed to load classifier package: {e}") from e

paradigma/config.py

@@ -0,0 +1,314 @@
+from typing import Dict, List
+from paradigma.constants import DataColumns, DataUnits
+import numpy as np
+
+class BaseConfig:
+    def __init__(self) -> None:
+        self.meta_filename = ''
+        self.values_filename = ''
+        self.time_filename = ''
+
+    def set_sensor(self, sensor: str) -> None:
+        """Sets the sensor and derived filenames"""
+        self.sensor: str = sensor
+        self.set_filenames(sensor)
+
+    def set_filenames(self, prefix: str) -> None:
+        """Sets the filenames based on the prefix. This method is duplicated from `gaits_analysis_config.py`.
+        
+        Parameters
+        ----------
+        prefix : str
+            The prefix for the filenames.
+        """
+        self.meta_filename = f"{prefix}_meta.json"
+        self.time_filename = f"{prefix}_time.bin"
+        self.values_filename = f"{prefix}_values.bin"
+
+class IMUConfig(BaseConfig):
+
+    def __init__(self) -> None:
+        super().__init__()
+
+        self.set_filenames('IMU')
+
+        self.acceleration_units = DataUnits.ACCELERATION
+        self.rotation_units = DataUnits.ROTATION
+
+        self.axes = ["x", "y", "z"]
+
+        self.accelerometer_cols: List[str] = [
+            DataColumns.ACCELEROMETER_X,
+            DataColumns.ACCELEROMETER_Y,
+            DataColumns.ACCELEROMETER_Z,
+        ]
+        self.gyroscope_cols: List[str] = [
+            DataColumns.GYROSCOPE_X,
+            DataColumns.GYROSCOPE_Y,
+            DataColumns.GYROSCOPE_Z,
+        ]
+        self.gravity_cols: List[str] = [
+            DataColumns.GRAV_ACCELEROMETER_X,
+            DataColumns.GRAV_ACCELEROMETER_Y,
+            DataColumns.GRAV_ACCELEROMETER_Z,
+        ]
+
+        self.d_channels_accelerometer = {
+            DataColumns.ACCELEROMETER_X: self.acceleration_units,
+            DataColumns.ACCELEROMETER_Y: self.acceleration_units,
+            DataColumns.ACCELEROMETER_Z: self.acceleration_units,
+        }
+        self.d_channels_gyroscope = {
+            DataColumns.GYROSCOPE_X: self.rotation_units,
+            DataColumns.GYROSCOPE_Y: self.rotation_units,
+            DataColumns.GYROSCOPE_Z: self.rotation_units,
+        }
+        self.d_channels_imu = {**self.d_channels_accelerometer, **self.d_channels_gyroscope}
+
+        self.sampling_frequency = 100
+        self.lower_cutoff_frequency = 0.2
+        self.upper_cutoff_frequency = 3.5
+        self.filter_order = 4
+
+
+class PPGConfig(BaseConfig):
+
+    def __init__(self) -> None:
+        super().__init__()
+
+        self.set_filenames('PPG')
+
+        self.ppg_colname = DataColumns.PPG
+
+        self.sampling_frequency = 30
+        self.lower_cutoff_frequency = 0.4
+        self.upper_cutoff_frequency = 3.5
+        self.filter_order = 4
+
+        self.d_channels_ppg = {
+            DataColumns.PPG: DataUnits.NONE
+        }
+
+
+# Domain base configs
+class GaitConfig(IMUConfig):
+
+    def __init__(self, step) -> None:
+        super().__init__()
+
+        self.set_sensor('accelerometer')
+
+        # ----------
+        # Segmenting
+        # ----------
+        self.max_segment_gap_s = 1.5
+        self.min_segment_length_s = 1.5
+
+        if step == 'gait':
+            self.window_length_s: float = 6
+            self.window_step_length_s: float = 1
+        else:
+            self.window_length_s: float = 3
+            self.window_step_length_s: float = self.window_length_s * 0.25
+
+        # -----------------
+        # Feature extraction
+        # -----------------
+        self.window_type: str = "hann"
+        self.spectrum_low_frequency: int = 0
+        self.spectrum_high_frequency: int = int(self.sampling_frequency / 2)
+
+        # Power in specified frequency bands
+        self.d_frequency_bandwidths: Dict[str, List[float]] = {
+            "power_below_gait": [0.2, 0.7],
+            "power_gait": [0.7, 3.5],
+            "power_tremor": [3.5, 8],
+            "power_above_tremor": [8, 25],
+        }
+
+        # Mel frequency cepstral coefficients
+        self.mfcc_low_frequency: int = 0
+        self.mfcc_high_frequency: int = 25
+        self.mfcc_n_dct_filters: int = 15
+        self.mfcc_n_coefficients: int = 12
+
+        # Dominant frequency of first harmonic of arm swing
+        self.angle_fmin: float = 0.5
+        self.angle_fmax: float = 1.5
+
+        # -----------------
+        # TSDF data storage
+        # -----------------
+        self.d_channels_values: Dict[str, str] = {
+            "accelerometer_std_norm": DataUnits.GRAVITY,
+            "accelerometer_x_grav_mean": DataUnits.GRAVITY,
+            "accelerometer_y_grav_mean": DataUnits.GRAVITY,
+            "accelerometer_z_grav_mean": DataUnits.GRAVITY,
+            "accelerometer_x_grav_std": DataUnits.GRAVITY,
+            "accelerometer_y_grav_std": DataUnits.GRAVITY,
+            "accelerometer_z_grav_std": DataUnits.GRAVITY,
+            "accelerometer_x_power_below_gait": DataUnits.POWER_SPECTRAL_DENSITY,
+            "accelerometer_y_power_below_gait": DataUnits.POWER_SPECTRAL_DENSITY,
+            "accelerometer_z_power_below_gait": DataUnits.POWER_SPECTRAL_DENSITY,
+            "accelerometer_x_power_gait": DataUnits.POWER_SPECTRAL_DENSITY,
+            "accelerometer_y_power_gait": DataUnits.POWER_SPECTRAL_DENSITY,
+            "accelerometer_z_power_gait": DataUnits.POWER_SPECTRAL_DENSITY,
+            "accelerometer_x_power_tremor": DataUnits.POWER_SPECTRAL_DENSITY,
+            "accelerometer_y_power_tremor": DataUnits.POWER_SPECTRAL_DENSITY,
+            "accelerometer_z_power_tremor": DataUnits.POWER_SPECTRAL_DENSITY,
+            "accelerometer_x_power_above_tremor": DataUnits.POWER_SPECTRAL_DENSITY,
+            "accelerometer_y_power_above_tremor": DataUnits.POWER_SPECTRAL_DENSITY,
+            "accelerometer_z_power_above_tremor": DataUnits.POWER_SPECTRAL_DENSITY,
+            "accelerometer_x_dominant_frequency": DataUnits.FREQUENCY,
+            "accelerometer_y_dominant_frequency": DataUnits.FREQUENCY,
+            "accelerometer_z_dominant_frequency": DataUnits.FREQUENCY,
+        }
+
+        for mfcc_coef in range(1, self.mfcc_n_coefficients + 1):
+            self.d_channels_values[f"accelerometer_mfcc_{mfcc_coef}"] = DataUnits.GRAVITY
+
+        if step == 'arm_activity':
+            for mfcc_coef in range(1, self.mfcc_n_coefficients + 1):
+                self.d_channels_values[f"gyroscope_mfcc_{mfcc_coef}"] = DataUnits.GRAVITY
+
+
+class TremorConfig(IMUConfig):
+
+    def __init__(self, step: str | None = None) -> None:
+        """
+        Parameters
+        ----------
+        step : str (optional)
+            The step of the tremor pipeline. Can be 'features' or 'classification'.
+        """
+        super().__init__()
+
+        self.set_sensor('gyroscope')
+
+        # ----------
+        # Segmenting
+        # ----------
+        self.window_length_s: float = 4
+        self.window_step_length_s: float = 4
+
+        # -----------------
+        # Feature extraction
+        # -----------------
+        self.window_type = 'hann'
+
+        # Power spectral density
+        self.overlap_fraction: float = 0.8
+        self.segment_length_s: float = 3
+        self.spectral_resolution: float = 0.25
+        self.fmin_peak_search: float = 1
+        self.fmax_peak_search: float = 25
+        self.fmin_below_rest_tremor: float = 0.5
+        self.fmax_below_rest_tremor: float = 3
+        self.fmin_rest_tremor: float = 3
+        self.fmax_rest_tremor: float = 7
+
+        # Mel frequency cepstral coefficients
+        self.fmin_mfcc: float = 0
+        self.fmax_mfcc: float = 25
+        self.n_dct_filters_mfcc: int = 15
+        self.n_coefficients_mfcc: int = 12
+
+        # --------------
+        # Classification
+        # --------------
+        self.movement_threshold: float = 50
+
+        # -----------
+        # Aggregation
+        # -----------
+        self.aggregates_tremor_power: List[str] = ['mode', 'median', '90p']
+
+        # -----------------
+        # TSDF data storage
+        # -----------------
+        if step == 'features':
+            self.d_channels_values: Dict[str, str] = {}
+            for mfcc_coef in range(1, self.n_coefficients_mfcc + 1):
+                self.d_channels_values[f"mfcc_{mfcc_coef}"] = "unitless"
+
+            self.d_channels_values["freq_peak"] = "Hz"
+            self.d_channels_values["below_tremor_power"] = "(deg/s)^2"
+            self.d_channels_values["tremor_power"] = "(deg/s)^2"
+        elif step == 'classification':
+            self.d_channels_values = {
+                DataColumns.PRED_TREMOR_PROBA: "probability",
+                DataColumns.PRED_TREMOR_LOGREG: "boolean",
+                DataColumns.PRED_TREMOR_CHECKED: "boolean",
+                DataColumns.PRED_ARM_AT_REST: "boolean"
+            }
+
+        
+class HeartRateConfig(PPGConfig):
+    def __init__(self, sensor: str = 'ppg', min_window_length_s: int = 30) -> None:
+        super().__init__()
+
+        # ----------
+        # Segmenting
+        # ----------
+        self.window_length_s: int = 6
+        self.window_step_length_s: int = 1
+        self.window_overlap_s = self.window_length_s - self.window_step_length_s
+
+        self.accelerometer_cols = IMUConfig().accelerometer_cols
+
+        # -----------------------
+        # Signal quality analysis
+        # -----------------------
+        self.freq_band_physio = [0.75, 3] # Hz
+        self.bandwidth = 0.2   # Hz      
+        self.freq_bin_resolution = 0.05 # Hz
+
+        # ---------------------
+        # Heart rate estimation
+        # ---------------------
+        self.set_tfd_length(min_window_length_s)  # Set tfd length to default of 30 seconds
+        self.threshold_sqa = 0.5
+        self.threshold_sqa_accelerometer = 0.13
+
+        hr_est_length = 2
+        self.hr_est_samples = hr_est_length * self.sampling_frequency
+
+        # Time-frequency distribution parameters
+        self.kern_type = 'sep'
+        win_type_doppler = 'hamm'
+        win_type_lag = 'hamm'
+        win_length_doppler = 8
+        win_length_lag = 1
+        doppler_samples = self.sampling_frequency * win_length_doppler
+        lag_samples = win_length_lag * self.sampling_frequency
+        self.kern_params = {
+            'doppler': {
+                'win_length': doppler_samples,
+                'win_type': win_type_doppler,
+            },
+            'lag': {
+                'win_length': lag_samples,
+                'win_type': win_type_lag,
+            }
+        }
+
+        self.set_sensor(sensor)
+
+    def set_tfd_length(self, tfd_length: int):
+        self.tfd_length = tfd_length
+        self.min_hr_samples = int(round(self.tfd_length * self.sampling_frequency))
+
+    def set_sensor(self, sensor):
+        self.sensor = sensor 
+
+        if sensor not in ['ppg', 'imu']:
+            raise ValueError(f"Invalid sensor type: {sensor}")
+        
+        if sensor == 'imu':
+            self.sampling_frequency = IMUConfig().sampling_frequency
+        else:
+            self.sampling_frequency = PPGConfig().sampling_frequency
+
+        self.window_length_welch = 3 * self.sampling_frequency
+        self.overlap_welch_window = self.window_length_welch // 2
+        self.nfft = len(np.arange(0, self.sampling_frequency/2, self.freq_bin_resolution))*2

paradigma/constants.py

@@ -14,17 +14,52 @@ class DataColumns():
     GYROSCOPE_Z : str = "gyroscope_z"
     PPG : str = "green"
     TIME : str = "time"
+    SEGMENT_NR : str = "segment_nr"
+    SEGMENT_CAT: str = "segment_category"
 
-    # The following are used in gait analysis
-    GRAV_ACCELEROMETER_X : str = "grav_accelerometer_x"
-    GRAV_ACCELEROMETER_Y : str = "grav_accelerometer_y"
-    GRAV_ACCELEROMETER_Z : str = "grav_accelerometer_z"
+    # Gait 
+    GRAV_ACCELEROMETER_X : str = "accelerometer_x_grav"
+    GRAV_ACCELEROMETER_Y : str = "accelerometer_y_grav"
+    GRAV_ACCELEROMETER_Z : str = "accelerometer_z_grav"
+    PRED_GAIT_PROBA: str = "pred_gait_proba"
     PRED_GAIT : str = "pred_gait"
-    PRED_ARM_SWING : str = "pred_arm_swing"
+    PRED_NO_OTHER_ARM_ACTIVITY_PROBA: str = "pred_no_other_arm_activity_proba"
+    PRED_NO_OTHER_ARM_ACTIVITY : str = "pred_no_other_arm_activity"
     ANGLE : str = "angle"
-    ANGLE_SMOOTH : str = "angle_smooth"
     VELOCITY : str = "velocity"
-    SEGMENT_NR : str = "segment_nr"
+    DOMINANT_FREQUENCY: str = "dominant_frequency"
+    RANGE_OF_MOTION: str = "range_of_motion"
+    PEAK_VELOCITY: str = "peak_velocity"
+
+    # The following are used in tremor analysis
+    PRED_TREMOR_PROBA: str = "pred_tremor_proba"
+    PRED_TREMOR_LOGREG : str = "pred_tremor_logreg"
+    PRED_TREMOR_CHECKED : str = "pred_tremor_checked"
+    PRED_ARM_AT_REST: str = "pred_arm_at_rest"
+
+    # Constants for PPG features
+    VARIANCE: str = "variance"
+    MEAN: str = "mean"
+    MEDIAN: str = "median"
+    KURTOSIS: str = "kurtosis"
+    SKEWNESS: str = "skewness"
+    DOMINANT_FREQUENCY: str = "dominant_frequency"
+    RELATIVE_POWER: str = "relative_power"
+    SPECTRAL_ENTROPY: str = "spectral_entropy"
+    SIGNAL_NOISE_RATIO: str = "signal_noise_ratio"
+    SECOND_HIGHEST_PEAK: str = "second_highest_peak"
+    POWER_RATIO: str = "power_ratio"
+
+    # Constants for PPG SQA feature using accerometer data
+    ACC_POWER_RATIO: str = "acc_power_ratio"
+
+    # Constants for SQA classification
+    PRED_SQA_PROBA: str = "pred_sqa_proba"
+    PRED_SQA_ACC_LABEL: str = "pred_sqa_acc_label"
+    PRED_SQA: str = "pred_sqa"
+
+    # Constants for heart rate
+    HEART_RATE: str = "heart_rate"
     
 @dataclass(frozen=True)
 class DataUnits():
@@ -57,9 +92,15 @@ class TimeUnit():
     """
     RELATIVE_MS : str = "relative_ms"
     """ The time is relative to the start time in milliseconds. """
+    RELATIVE_S : str = "relative_s"
+    """ The time is relative to the start time in seconds. """
     ABSOLUTE_MS : str = "absolute_ms"
     """ The time is absolute in milliseconds. """
+    ABSOLUTE_S : str = "absolute_s"
+    """ The time is absolute in seconds. """
     DIFFERENCE_MS : str = "difference_ms"
     """ The time is the difference between consecutive samples in milliseconds. """
+    DIFFERENCE_S : str = "difference_s"
+    """ The time is the difference between consecutive samples in seconds. """
 
 UNIX_TICKS_MS: int = 1000