AID: Open-Source Anechoic Interferer Dataset

A dataset of anechoic recordings of various sound sources encountered in domestic environments is provided, which is intended to be a resource of non-stationary, environmental noise signals that, when convolved with acoustic impulse responses, can be used to simulate complex acoustic scenes. The dataset consists of anechoic recordings of 43 different types of sound sources encountered in domestic environments, with the number of individual recordings per sound source varying between two and eleven. The sound sources, which are mostly household devices and utilities, include door keys, plastic bags, clothing, a drilling machine, an electric blender, glass jars and metal boxes but also a few human-made sounds, such as clapping, breathing, snapping or whistling. The recordings cover a wide range of timbres. Multiple sounds from every individual source were recorded by different ways of excitation, such as hitting and shaking, or switching on and off the electric devices. Three different microphones were used to record the various sound sources. In addition, a Python library is provided that can be used to randomly arrange multiple anechoic noise recordings into a single channel interference signal. The number of individual recordings concurrently playing at any point in time in an interference signal can be specified by the user, providing control over the temporal density. The signal generator implementation is hosted on GitHub.本数据集收录了家庭环境中常见各类声源的消声录音（anechoic recordings），旨在作为非平稳环境噪声信号的资源库，将其与声学冲激响应（acoustic impulse response）卷积后，可用于模拟复杂声学场景。该数据集包含43种家庭环境常见声源的消声录音，每个声源对应的独立录音数量介于2至11条区间内。这些声源多为家用设备与日常用品，涵盖门钥匙、塑料袋、衣物、电钻、电动搅拌机、玻璃罐及金属盒等，同时包含少量人工发声样本，如拍手、呼吸、打响指或吹口哨。本次录制的录音覆盖了丰富的音色范围。针对每个独立声源，通过多种激发方式录制多段录音，例如敲击、摇晃，或开启、关停用电设备。本次录制共使用三款不同的麦克风。此外，本数据集还附带一款Python库，可用于将多条消声噪声录音随机组合为单通道干扰信号。用户可自行指定干扰信号中任意时刻同时播放的独立录音数量，从而实现对时域密度的调控。该信号生成器的实现代码托管于GitHub平台。