This dataset provides German speech samples augmented with realistic noise scenarios such as office noise, street noise, white noise, echo, and lowpass filtering.It is designed for testing and improving the robustness of ASR (Automatic Speech Recognition) systems.
Original voice samples: Mozilla Common Voice (CC0)
Noise layers: custom augmented (see noise.rolgor.de)
Please respect speaker… See the full description on the dataset page:
https://huggingface.co/datasets/rolgor/noise-dataset-de.
