Timbre analysis

Stroke Classification

Mridangam Stroke Classification

Note

REQUIRES: essentia

class compiam.timbre.stroke_classification.mridangam_stroke_classification.MridangamStrokeClassification

Mridangam stroke classification.

dict_strokes()

List and convert to indexed dict the available mridangam strokes in the dataset.

Returns:

dict with strokes as values and unique integer as keys.

list_strokes()

List available mridangam strokes in the dataset.

Returns:

list of strokes in the datasets.

load_mridangam_dataset(data_home=None, version='default', download=True)

Load mirdata dataloader for mirdangam stroke.

Parameters:

• data_home – folder where the dataset is found.

• version – version of the dataset to use.

• download – if True the dataset is downloaded.

Returns:

None, but initializes the dataset of the class and the file dict of strokes.

predict(file_list)

Predict stroke type from list of files.

Parameters:

file_list – list of files for prediction.

Returns:

dict containing filenames as keys and estimated strokes as values.

train_model(model_type='svm', load_computed=False, balance=False, balance_ref='random')

Train a support vector machine for stroke classification.

Parameters:

• model_type – type of model to train.

• model_type – bool to indicate if the features are computed or loaded from file.

• balance – balance the number of instances per class to prevent biases.

• balance_ref – reference class for data balancement.

Returns:

accuracy in percentage and rounded to two decimals