Data and code for "Passive acoustic computing with cochlea-inspired tonotopic piezoelectric arrays for temporal speech-command recognition"

This record contains source-data records, complete frozen recognition inputs, analysis code, and representative signal examples for the accompanying study of cochlea-inspired tonotopic piezoelectric arrays. The recognition inputs cover 619 multichannel array samples and 588 microphone samples from five anonymous participants and eight commands. Three routes are represented: CITPA-24 with temporal features (A), microphone with spectral features (B), and microphone with the same temporal framework (C). The deposit includes anonymous metadata, fixed evaluation splits, prediction records, enrollment outcomes, backend cost measurements, 15 P0 models, and 15 inference-only model bundles. For easier transfer, the recognition inputs are distributed across eight smaller archives. The supplied restoration utility reconstructs the original feature file and verifies all 44 recognition input files. No recognition samples or numerical inputs were removed. Download and execution instructions are provided in README.md. Full continuous acquisition records and additional instrument-native files may be requested from He Tian ([e-mail]).