A Deep Learning-based method for Multi-Zone Sound Field Synthesis

Code repository for my master thesis on A Deep Learning-based method for Multi-Zone Sound Field Synthesis.

• Dependencies
• Data Generation
• Network Training
• Results Computation

Dependencies

• Python, has been tested with version 3.6.9
• Numpy, sci-kit-image, sci-kit-learn, tqdm, matplotlib
• Jax
• Tensorflow 2.4.1
• sfs

Data generation

There are two different scripts to generate the data, generate_test_data_linear_array.py and generate_train_data_linear_array.py. The parameters used for data generation (e.g. sources position, array position) are defined into data_lib/params_linear.

The command-line arguments are the following

• gt_soundfield: bool, True if want to generate also data related to ground truth sound field
• dataset_path: String, folder where to store the dataset

Network training

There is a script to generate the data with the linear array, namely generate_data_linear_array.py. The parameters used for data generation (e.g. sources position, array position) are defined into data_lib/params_linear.

• epochs: Int, number of epochs
• batch_size: Int, dimension of batches used for training
• log_dir: String, folder where store training logs accessible via Tensorboard
• gt_soundfield_dataset_path: String, path to numpy array where ground truth sound-field data are contained
• learning_rate: Float, learning rate used for the optimizer
• green_function: String, path to numpy array where green function between secondary sources and evaluation points is contained

Results computation

To compute Reproduction Error (MSE) and Structural Similarity Index (SSIM), run the code into plot_results.py with the following arguments

• dataset_path: String, path to the folder to save results stored into arrays.
• array_type: String, type of array, i.e. linear or circular

N.B. pre-trained models used to compute the results can be found in folder _models_s