mouse-telemetry

mouse-telemetry

DISCLAIMER: This is intended for benevolent analytics and machine learning tasks; Do NOT use this for nefarious purposes

High-fidelity mouse telemetry collector for Rust. Captures cursor movement, clicks, and scrolls; computes derived kinematics (velocity, acceleration, jerk, smoothness, Fitts-style indices); and writes everything to CSV for analysis.

Highlights

• Event capture: Cursor moves, left/middle/right clicks, wheel up/down
• Derived metrics: Velocity, acceleration, jerk, momentum, smoothness, path deviation, idle time, simple Fitts metrics
• CSV export: Human-readable schema with a stable header
• Runtime tuning: Smoothing and derivative clamping controls
• Live stats: Summary strings for easy HUD/overlay display

Quickstart

You can run the collector in two ways:

1) Spawn an internal listener window

The library will open a minimal window and record mouse activity within it. A left-click marks the end of a batch; batches are processed and appended to CSV.

use std::time::Duration;
use std::sync::Arc;
use mouse_collector::mouse_collector::MouseCollector;

#[tokio::main]
async fn main() {
    // Start immediately; disable CLI reporting (stats available via APIs below)
    let collector: Arc<MouseCollector> = MouseCollector::new(true, false).await;

    // Optional: change where CSV is written
    collector.set_data_dir("data".to_string());

    // Keep the app alive; interact with the spawned window
    // Each left-click will finalize the current batch and write rows to CSV.
    loop { tokio::time::sleep(Duration::from_secs(1)).await; }
}

Notes:

• The listener records events that occur within the spawned window (focus matters).
• A left-click finalizes the current batch and triggers CSV writes.

2) Integrate with your own app (push events programmatically)

Use your own event source, push events to the collector, and run only the recorder thread.

use std::{sync::Arc, thread, time::Duration};
use mouse_collector::mouse_collector::MouseCollector;

#[tokio::main]
async fn main() {
    let collector: Arc<MouseCollector> = MouseCollector::new(false, false).await;

    // Write output to a custom directory/filename
    collector.set_data_dir("data".to_string());
    if let Ok(mut name) = collector.record_filename.write() {
        *name = "session_01.csv".to_string();
    }

    // Start only the CSV recording thread
    let _recorder = collector.start_recording_thread();

    // Feed your own events
    collector.push_move((100.0, 100.0));
    collector.push_move((120.0, 110.0));
    collector.push_move((180.0, 140.0));

    // A left-click marks the end of the current batch and triggers CSV write
    collector.push_left_click((180.0, 140.0));

    thread::sleep(Duration::from_millis(500));
}

Live stats (for overlays/HUDs)

• get_latest_derived_summary() → single compact string
• get_latest_derived_stats() → Vec for column/row UI layouts

let summary = collector.get_latest_derived_summary();
let stats = collector.get_latest_derived_stats();
println!("{summary}");
for s in stats { println!("{s}"); }

Configuration

Output location

• Default directory: data/
• Default filename: mouse_data_YYYYMMDD_HHMMSS.csv (auto-increments if the file exists)

collector.set_data_dir("my_data".to_string());
if let Ok(mut name) = collector.record_filename.write() {
    *name = "experiment_a.csv".to_string();
}

Smoothing and derivative clamp

These influence stability of velocity/acceleration estimates:

// Clamp very small dt (seconds) to avoid huge derivatives (default ~0.004)
collector.set_min_derivative_dt(0.003);

// Exponential moving average alpha in (0, 1]; higher = more responsive
collector.set_smoothing_alpha(0.35);

How batching and writing work

• Events are buffered in memory.
• A left-click acts as a delimiter that closes the current batch.
• When a batch is closed and has >1 event, the collector:
  • computes kinematics and derived metrics
  • appends rows to the CSV file

Tip: If you drive the collector programmatically, ensure you call push_left_click periodically to persist data in well-defined segments (e.g., per task, per gesture, per trial).

CSV schema

All numeric values are floating-point unless noted. Units: time in seconds, positions in pixels, speeds in pixels/second.

Platform notes

• Uses winit for portable window/input handling (Linux, Windows, macOS).
• Listener mode records events within the library’s window. For global OS-level hooks, integrate with your own input source and push events via push_move / push_left_click.

Troubleshooting

• No rows appear in the CSV: Ensure a left-click occurs to close a batch. If you push events programmatically, call push_left_click to finalize.
• No CSV file created: Verify the data directory exists or call set_data_dir. Also confirm your process has write permissions.
• Stats look noisy: Lower set_smoothing_alpha or increase set_min_derivative_dt.
• Nothing is recorded in listener mode: Make sure the spawned window is focused and your cursor is inside it.

License

MIT