Real-template benchmark data for resolving superimposed action potentials

Résumé fourni par la source

This dataset contains real-template benchmark sets and working-example metadata used for evaluating a candidate-refined convex-lifted nonnegative least-squares method for resolving superimposed action potentials. The data support the associated Technical Note on resolving superimposed potentials in intramuscular electromyographic decomposition and neural spike sorting.The waveform morphologies are experimentally derived action-potential templates extracted from recorded multiunit signals. They are not analytic synthetic pulses. In the associated experiments, only the occurrence times, amplitude gains, and noise conditions were simulated so that exact ground-truth shifts were available for objective algorithmic validation. The same superposition model applies to overlapping motor-unit action potentials in intramuscular EMG decomposition and overlapping neural spikes in spike sorting.The package includes three 128-sample real-template benchmark sheets. set1 contains 10 templates and is used as a wide-dynamic-range benchmark. set2 contains 6 templates and is used as a lower-cardinality, low-to-moderate-amplitude subset. set3 contains 9 templates and is used as a higher-energy, correlation-rich action-potential benchmark. The files are provided in multiple formats to support reuse in Python, MATLAB, spreadsheet software, and other signal-processing environments. Working-example files specify active template identities, true integer shifts, amplitude gains, SNR conditions, and descriptions for deterministic examples used in the associated reproducibility package.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.