Skip to content

[BUG] Incorrect computation of unsigned AUC (absolute area under the curve) #180

Description

@dmfolgado

The current implementation of the auc feature applies the absolute value after summing consecutive samples:
np.sum(0.5 * np.diff(t) * np.abs(np.array(signal[:-1]) + np.array(signal[1:])))

This effectively computes (abs(y_i + y_{i+1}) / 2) * Δt, which is not equivalent to the correct trapezoidal rule for the unsigned area ( (abs(y_i) + abs(y_{i+1})) / 2 ) * Δt.

When the signal crosses zero, the current implementation underestimates the area. For example, for a 1 Hz signal = [3, -3], the current implementation gives 0.0 instead of the correct 3.0.

The unsigned AUC should integrate the absolute value of each sample, not the absolute value of their sum. We should replace the current implementation by the SciPy trapezoidal rule implementation, which is also faster.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

bugSomething isn't workingenhancementNew feature or request

Type

No type

Projects

No projects

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions