Part 2: Recording Interface

The recording interface is the second component of a visual interface in MyoGestic.

It is responsible for managing the ground-truth data collection. Since the ground-truth data depends on the visual interface, the recording interface must define the data collection process.

Step 1: Create the setup UI

You will need an UI that will allow the user to start the recording, visualize the progress, and save the data.

Note

To start copy the UI files in myogestic > gui > widgets > visual_interfaces > ui and adapt them with the functionality you need.

You need to modify them with QT-Designer and convert them using UIC to a python file.

Step 2: Understand what is needed in the recording interface

Note

A recording interface is a class that inherits from

RecordingInterfaceTemplate(main_window[, ...])

Base class for the recording interface of a visual interface.

Please read the documentation of the class and make a mental note of what you have to provide (e.g. signals, methods, attributes) and what you have to implement (e.g. start_recording, stop_recording, accept_recording).

Step 3: Implement a recording interface (Example Virtual Hand Interface)

This example focuses on implementing and adding the recording interface for the Virtual Hand Interface using the VirtualHandInterface_RecordingInterface class from recording_interface.py. We explain how it is constructed and registered into MyoGestic via CONFIG_REGISTRY in config.py.

Steps:

1. Class Initialization and UI Setup

2. Starting & Stopping Recordings

3. Managing Recording Sessions

4. Resetting the Interface

Step 3.1: Class Initialization and UI Setup

This step initializes the recording interface’s buffers and task-related attributes. It also configures the UI elements required for this interface.

Imports

import time

import numpy as np
from PySide6.QtCore import SignalInstance
from PySide6.QtGui import QCloseEvent

from myogestic.gui.widgets.logger import LoggerLevel
from myogestic.gui.widgets.templates.visual_interface import RecordingInterfaceTemplate
from myogestic.gui.widgets.visual_interfaces.virtual_hand_interface.ui import (
    Ui_RecordingVirtualHandInterface,
)
from myogestic.utils.constants import RECORDING_DIR_PATH

KINEMATICS_SAMPLING_FREQUENCY = 60

Class Initialization

class VirtualHandInterface_RecordingInterface(RecordingInterfaceTemplate):
    """
    Class for the recording interface of the Virtual Hand Interface.

    This class is responsible for handling the recording of EMG and kinematics data.

    Parameters
    ----------
    main_window : MainWindow
        The main window of the application.
    name : str
        The name of the interface, by default "VirtualHandInterface".

        .. important:: This name is used to identify the interface in the main window. It should be unique.
    incoming_message_signal : SignalInstance
        The signal instance used to receive incoming messages from the device.
    """

    def __init__(
        self,
        main_window,
        name: str = "VirtualHandInterface",
        incoming_message_signal: SignalInstance = None,
    ) -> None:
        super().__init__(
            main_window,
            name,
            ui=Ui_RecordingVirtualHandInterface(),
            incoming_message_signal=incoming_message_signal,
            ground_truth__nr_of_recording_values=9,
            ground_truth__task_map={
                "rest": 0,
                "index": 1,
                "thumb": 2,
                "middle": 3,
                "ring": 4,
                "pinky": 5,
                "fist": 6,
                "pinch": 7,
                "3fpinch": 8,
                "pointing": 9,
            }
        )

        RECORDING_DIR_PATH.mkdir(parents=True, exist_ok=True)

        self._current_task: str = ""
        self._kinematics__buffer = []

        self._recording_protocol = self._main_window.protocols[0]

        self._has_finished_kinematics: bool = False
        self._start_time: float = 0

        self.initialize_ui_logic()

UI Setup

    def initialize_ui_logic(self) -> None:
        """Initializes the logic for the UI elements."""
        ui: Ui_RecordingVirtualHandInterface = self.ui

        self._main_window.ui.recordVerticalLayout.addWidget(ui.recordRecordingGroupBox)
        self._main_window.ui.recordVerticalLayout.addWidget(
            ui.recordReviewRecordingStackedWidget
        )

        self.record_group_box = ui.recordRecordingGroupBox
        self.record_task_combo_box = ui.recordTaskComboBox
        self.record_duration_spin_box = ui.recordDurationSpinBox
        self.record_toggle_push_button = ui.recordRecordPushButton

        self.review_recording_stacked_widget = ui.recordReviewRecordingStackedWidget
        self.review_recording_task_label = ui.reviewRecordingTaskLabel
        self.review_recording_label_line_edit = ui.reviewRecordingLabelLineEdit

        self.review_recording_accept_push_button = ui.reviewRecordingAcceptPushButton
        self.review_recording_reject_push_button = ui.reviewRecordingRejectPushButton

        self.use_kinematics_check_box = ui.recordUseKinematicsCheckBox

        self.record_toggle_push_button.toggled.connect(self.start_recording)
        self.review_recording_accept_push_button.clicked.connect(self.accept_recording)
        self.review_recording_reject_push_button.clicked.connect(self.reject_recording)

        self.record_ground_truth_progress_bar.setValue(0)
        self.review_recording_stacked_widget.setCurrentIndex(0)

Step 3.2: Starting & Stopping Recordings

This step manages the logic for starting or stopping recording sessions, which includes preparing recording parameters and updating the UI.

Start Recording Preparation - This method prepares the recording session.

    def start_recording_preparation(self) -> bool:
        """Prepares the recording process by checking if the device is streaming."""
        if (
            not self._main_window.device__widget._get_current_widget()._device._is_streaming
        ):
            self._main_window.logger.print(
                "Biosignal device not streaming!", level=LoggerLevel.ERROR
            )
            return False

        self.kinematics_recording_time = int(
            self.record_duration_spin_box.value() * KINEMATICS_SAMPLING_FREQUENCY
        )
        self._kinematics__buffer = []
        return True

Start Recording - This method starts the recording session.

    def start_recording(self, checked: bool) -> None:
        """Starts the recording process."""
        if checked:
            if not self.start_recording_preparation():
                self.record_toggle_push_button.setChecked(False)
                return

            if not self._recording_protocol.start_recording_preparation(
                self.record_duration_spin_box.value()
            ):
                self.record_toggle_push_button.setChecked(False)
                return

            self._start_time = time.time()

            self.record_toggle_push_button.setText("Recording...")
            self.record_group_box.setEnabled(False)
            self._current_task = self.record_task_combo_box.currentText()

            if self.use_kinematics_check_box.isChecked():
                self.incoming_message_signal.connect(self.update_ground_truth_buffer)

            self._has_finished_kinematics = (
                not self.use_kinematics_check_box.isChecked()
            )

Step 3.3: Managing Recording Sessions

This step manages the recording session, which includes updating the UI and recording data.

Update Ground Truth Buffer - This method updates the ground-truth buffer with the current data.

    def update_ground_truth_buffer(self, data: np.ndarray) -> None:
        """Updates the buffer with the incoming kinematics data."""
        if not self.use_kinematics_check_box.isChecked():
            return

        self._kinematics__buffer.append((time.time(), data))
        current_samples = len(self._kinematics__buffer)
        self._set_progress_bar(
            self.record_ground_truth_progress_bar,
            current_samples,
            self.kinematics_recording_time,
        )

        if current_samples >= self.kinematics_recording_time:
            self._main_window.logger.print(
                f"Kinematics recording finished at: {round(time.time() - self._start_time)} seconds"
            )
            self._has_finished_kinematics = True
            self.incoming_message_signal.disconnect(self.update_ground_truth_buffer)
            self.check_recording_completion()

Check Recording Completion - This method checks if the recording session is complete.

    def check_recording_completion(self) -> None:
        """Checks if the recording process is complete and finishes it if so."""
        if (
            self._recording_protocol.is_biosignal_recording_complete
            and self._has_finished_kinematics
        ):
            self.finish_recording()

Finish Recording - This method finalizes the recording session.

    def finish_recording(self) -> None:
        """Finishes the recording process and switches to the review recording interface."""
        self.review_recording_stacked_widget.setCurrentIndex(1)
        self.record_toggle_push_button.setText("Finished Recording")
        self.review_recording_task_label.setText(self._current_task.capitalize())

Accept Recording - This method accepts the recorded data and saves it.

    def accept_recording(self) -> None:
        """
        Accepts the current recording and saves the data to a pickle file.

        The saved data is a dictionary containing:

        - emg: A 2D NumPy array of EMG signals with time samples as rows and channels as columns.
        - kinematics: A 2D NumPy array of kinematics data (empty if not used).
        - timings_emg: A 1D NumPy array of timestamps for EMG samples.
        - timings_kinematics: A 1D NumPy array of timestamps for kinematics samples (empty if not used).
        - label: The user-provided label for the recording.
        - task: The task being recorded.
        - device: The name of the device used for recording.
        - bad_channels: A list of channels marked as "bad."
        - _sampling_frequency: The EMG sampling frequency.
        - kinematics_sampling_frequency: The kinematics sampling frequency.
        - recording_time: The recording duration in seconds.
        - use_kinematics: Boolean indicating whether kinematics data was recorded.
        """
        label = self.review_recording_label_line_edit.text() or "default"
        (
            biosignal_data,
            biosignal_timings,
        ) = self._recording_protocol.retrieve_recorded_data()

        self.save_recording(
            biosignal=biosignal_data,
            biosignal_timings=biosignal_timings,
            ground_truth=(
                np.vstack([data for _, data in self._kinematics__buffer]).T
                if self.use_kinematics_check_box.isChecked()
                else np.array([])
            ),
            ground_truth_timings=(
                np.array([time_stamp for time_stamp, _ in self._kinematics__buffer])
                if self.use_kinematics_check_box.isChecked()
                else np.array([])
            ),
            recording_label=label,
            task=self._current_task,
            ground_truth_sampling_frequency=KINEMATICS_SAMPLING_FREQUENCY,
            use_as_classification=not self.use_kinematics_check_box.isChecked(),
            record_duration=self.record_duration_spin_box.value(),
        )

        self.reset_ui()
        self._main_window.logger.print(
            f"Recording of task {self._current_task.lower()} with label {label} accepted!"
        )

Reject Recording - This method rejects the recorded data and discards it.

    def reject_recording(self) -> None:
        """Rejects the current recording and resets the recording interface."""
        self.reset_ui()
        self._main_window.logger.print("Recording rejected.")

Step 3.4: Resetting the Interface

This step resets the recording interface to its initial state.

Reset UI - This method resets the recording interface’s UI elements.

    def reset_ui(self) -> None:
        """Resets the recording interface UI elements."""
        self.review_recording_stacked_widget.setCurrentIndex(0)
        self.record_toggle_push_button.setText("Start Recording")
        self.record_toggle_push_button.setChecked(False)
        self.record_group_box.setEnabled(True)

        self._recording_protocol._reset_recording_ui()

        self.record_ground_truth_progress_bar.setValue(0)
        self._kinematics__buffer.clear()

Close Event - This method handles the closing event of the recording interface.

    def close_event(self, _: QCloseEvent) -> None:
        """Closes the recording interface."""
        self.record_toggle_push_button.setChecked(False)
        self.reset_ui()
        self._recording_protocol.close_event(_)
        self._main_window.logger.print("Recording interface closed.")

Enable - This method enables the recording interface.

    def enable(self):
        """Enable the UI elements."""
        self.ui.recordRecordingGroupBox.setEnabled(True)
        self.ui.recordReviewRecordingStackedWidget.setEnabled(True)

Disable - This method disables the recording interface.

    def disable(self):
        """Disable the UI elements."""
        self.ui.recordRecordingGroupBox.setEnabled(False)
        self.ui.recordReviewRecordingStackedWidget.setEnabled(False)

Estimated memory usage: 528 MB