Getting Started

PyEyesWeb is an open-source Python library for the analysis of expressive qualities in human movement. It provides algorithms to extract movement features from motion data, enabling researchers, artists, and developers to quantify and analyze movement expressivity.

Installation

You can install PyEyesWeb using pip. Open your terminal and run:

pip install pyeyesweb

Quick Start

Here's a simple example to get you started with PyEyesWeb. This example demonstrates how to compute the smoothness index from motion data.

import csv
from pyeyesweb.data_models import SlidingWindow
from pyeyesweb.low_level import Smoothness

# 1. Initialize the Smoothness analyzer
smoothness = Smoothness(rate_hz=50.0)

# 2. Create a sliding window to store the last 100 frames of data
window = SlidingWindow(max_length=100, n_columns=1)

# 3. Simulate a real-time loop reading from a CSV
with open('velocity_data.csv', 'r') as f:
    reader = csv.reader(f)

    for row in reader:
        # Extract the velocity module (assuming it is in the first column)
        velocity_val = float(row[0])

        # Append the new frame to the sliding window
        window.append([velocity_val]) #(1)!

        # Compute smoothness features on the current window
        sparc, jerk = smoothness(window) #(2)!

        # Check if a valid result was returned 
        # (feature may return None if window has not enough samples)
        if sparc is not None and jerk is not None:
            print(f"SPARC: {sparc:.3f} | Jerk: {jerk:.3f}")

1. The SlidingWindow expects a list of values for every frame (even if there is only 1 sample). As the loop runs, new data is added to the end, and old data is automatically discarded once the max_length is reached.

2. The smoothness callable processes the current state of the window. It returns the SPARC (Spectral Arc Length) and Jerk RMS. If the window does not yet contain enough data to compute the metric, it may return None.

Subpackages

PyEyesWeb is organized into subpackages analyzing movement features at different levels of abstraction and time scales ¹.

References