Audio Analysis Tool

The tool analyzes audio signals to produce a structured feature set that supports comparison and research. It applies FFT-based spectral analysis to derive centroid, bandwidth, rolloff, and spectral flux; it computes temporal features such as duration and tempo via onset detection; it performs loudness normalization and extracts MFCCs and chroma patterns. Each input file yields a feature vector and a list of feature names, plus per-file statistics (mean, std, min, max) to enable reproducibility across studies.
The system supports batch processing of multiple files, enabling a single analysis run across large datasets and consistent outputs for downstream modeling. Settings are configurable: window_size, overlap, target_sample_rate, and optional resampling. Results are deterministic, and the tool handles mono and stereo inputs, downsampling or upsampling when enabled.
Outputs can be exported in JSON, CSV, or YAML to integrate with data pipelines, notebooks, or reporting tools. This design serves researchers, educators, audio engineers, product testers, and QA teams who require objective metrics, scalable analysis, and fast comparisons without manual feature extraction. The approach emphasizes transparency in feature definitions, auditability of results, and compatibility with common analysis stacks.