# dereverberation-directional-feature

**Repository Path**: jinquan_wang/dereverberation-directional-feature

## Basic Information

- **Project Name**: dereverberation-directional-feature
- **Description**: Repo for the ICA 2019 Paper: Dereverberation based on deep neural networks with directional feature from spherical microphone array recordings
- **Primary Language**: Unknown
- **License**: GPL-2.0
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 1
- **Created**: 2020-10-14
- **Last Updated**: 2020-12-19

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# U-Net-based Speech Dereverberation with directional feature from spherical microphone array recordings

An implementation of [this paper](https://sytronik.github.io/assets/ICA2019.pdf).

# create.py

`create.py` performs following procedure:

1. Calculate **anechoic spherical harmonic domain (SHD) signals** from speech sources and spherical Fourier transform basis $\mathbf Y_s$ (`Ys`).
2. Calculate 32-channel spherical microphone array recordings from speech sources, room impulse responses (RIRs), and the modified inverse of rigid sphere modal strength $b^{-1}_n(kr)$ (`bEQf`).
3. Calculate **reverberant SHD signals** from the result of 2.
4. Perform STFT signals.
5. Calculate directional features, one of spatially-averaged intensity vector (**SIV**) and direction vector (**DV**).
6. Save magnitude and phase of the STFT of the 0-th order SHD signals and directional features.

Read docstring of `create.py` for usage.



## main.py

`main.py` is used to train or test DNNs.

Read docstring of `main.py` for usage.



## Model

The DNN model is based on FusionNet (U-Net-like DNN). Refer to `model` directory.



## Evaluation Metrics

Source codes for PESQ, STOI, and fwSegSNR are in `matlab_lib` directory.

Frequency-domain SegSNR is implemented in `audio_utils.py`.