Sound and Music Analysis (SoMA) Group

The Semantic Audio Feature Extraction Dataset (SAFE-DB) is a database of semantically annotated music production metadata collected from sound engineers. It contains data derived from four specific audio effects: a dynamic range compressor, an overdrive distortion, a parametric equaliser, and an algorithmic reverb. Each dataset entry consists of a user's semantic descriptor (e.g., 'warm', 'bright'), audio features extracted before and after processing, the specific plug-in parameters used, and optional user demographic/contextual data (age, location, language, experience, music genre, and instrument).

JSAP (JavaScript Audio Processor) units are HTML5 audio processors designed for sites using the Web Audio API. They enable the creation of advanced audio units with functionality similar to desktop VST and AU processors. JSAP leverages browser-defined AudioUnits and script processors to implement Digital Signal Processing (DSP) in JavaScript.

The Web Audio Evaluation Tool (WAET) is a toolbox built using the Web Audio API that allows researchers to configure and run various perceptual listening tests directly in a web browser. It eliminates the need for proprietary software or specialized programming knowledge and provides capabilities for remote testing environments and in-browser analytics.

JS-Xtract is a light-weight JavaScript library used for audio feature extraction. It is designed to be agnostic of the data's type or origin, allowing developers to perform analysis on time-series data either in real-time (by incorporating the library into a callback function) or on offline data. It is particularly useful for building listening test environments and online audio processing tools using JavaScript and HTML5, avoiding the need for compiled languages like C or Java.

The SAFE Project provides a suite of DAW plug-ins designed to replace low-level parameter control with statistically-defined semantic terms. Users can type a descriptive term (e.g., 'warm' or 'bright') into a text box to automatically adjust plug-in parameters based on server-side data. Users can also save their own parameter settings associated with a semantic term to contribute to the community-driven dataset.

The SAFE dataset provides an open-source Web API for software integration, allowing developers to access and retrieve feature data from the dataset. The API framework is available for download and hosting via GitHub.

# Source code for the SAFE Web API
https://github.com/semanticaudio/safe-api

The Snare Drum Data Set (SDDS) is a comprehensive collection of 2,544 WAV audio files captured from 48 performances. It is designed for researchers needing extreme tonal variation of snare drums, mimicking real-world studio practices. The dataset includes recordings of varied velocity strikes using 10 tonally distinct snare drums, 53 different studio microphones, 3 mic preamps, and various dampening methods, utilizing both close and far micing positions in an acoustically treated studio.

The SAFE dataset of music production metadata is available as a set of RDF triples to ensure compatibility with semantic web technologies. This dataset can be accessed through a SPARQL endpoint or a semantic web API.

The Sound and Music Analysis (SoMA) Group provides an automatic drum transcription (ADT) library for the annotation of pitch and duration information from acoustic signals. This tool is part of their research into improving generalisability and peak peaking in ADT systems.

The Semantic Audio Feature Extraction Database (SAFE-DB) can be downloaded as a zip archive from the DMT Lab.

http://dmtlab.bcu.ac.uk/SAFE/data/all.zip

The source code and implementation examples for the Web Audio Evaluation Tool (WAET) are hosted on GitHub.