Functions — github.com/boozyguo/ClearWave

hub / github.com/boozyguo/ClearWave / functions

Functions37 in github.com/boozyguo/ClearWave

⨍ Functions 37 ◇ Types & classes 1

↓ 6 callersFunctioncreate_folder

(fd)

prepare_data.py:25

↓ 4 callersFunctioncalc_sp

Calculate spectrogram. Args: audio: 1darray. mode: string, 'magnitude' | 'complex' Returns: spectrogram: 2darray

prepare_data.py:338

↓ 4 callersFunctioneval

Validation function. Args: model: keras model. gen: object, data generator. x: 3darray, input, (n_segs, n_concat, n_freq

main_dnn.py:29

↓ 3 callersFunctionread_audio

(path, target_fs=None)

prepare_data.py:29

↓ 2 callersFunctioncalculate_pesq

Calculate PESQ of all enhaced speech. Args: workspace: str, path of workspace. speech_dir: str, path of clean speech. te

evaluate.py:88

↓ 2 callersFunctiondeframesig

Does overlap-add procedure to undo the action of framesig. Ref: From https://github.com/jameslyons/python_speech_features :param frames:

spectrogram_to_wave.py:85

↓ 2 callersMethodgenerate

(self, xs, ys)

data_generator.py:10

↓ 2 callersFunctionhalf_to_whole

Recover whole spectrogram from half spectrogram.

spectrogram_to_wave.py:57

↓ 2 callersFunctionifft_to_wav

Recover wav from whole spectrogram

spectrogram_to_wave.py:62

↓ 2 callersFunctionlog_sp

(x)

prepare_data.py:451

↓ 2 callersFunctionmat_2d_to_3d

Segment 2D array to 3D segments.

prepare_data.py:454

↓ 2 callersFunctionpad_or_trunc

(s, wav_len)

spectrogram_to_wave.py:66

↓ 2 callersFunctionpad_with_border

Pad the begin and finish of spectrogram with border frame value.

prepare_data.py:471

↓ 2 callersFunctionrms

Root mean square.

prepare_data.py:295

↓ 2 callersFunctionround_half_up

(number)

spectrogram_to_wave.py:118

↓ 2 callersFunctionwrite_audio

(path, audio, sample_rate)

prepare_data.py:40

↓ 1 callersFunctionadditive_mixing

Mix normalized source1 and source2. Args: s: ndarray, source1. n: ndarray, source2. Returns: mix_audio: ndarr

prepare_data.py:317

↓ 1 callersFunctioncalculate_mixture_features

Calculate spectrogram for mixed, speech and noise audio. Then write the features to disk. Args: workspace: str, path of workspace

prepare_data.py:125

↓ 1 callersFunctioncalculate_noisy_features

Calculate spectrogram for mixed, speech and noise audio. Then write the features to disk. Args: workspace: str, path of workspace.

prepare_data.py:232

↓ 1 callersFunctioncalculate_noisy_pesq

Calculate PESQ of all enhaced speech. Args: workspace: str, path of workspace. speech_dir: str, path of clean speech. te_snr: f

evaluate.py:55

↓ 1 callersFunctioncompute_scaler

Compute and write out scaler of data.

prepare_data.py:478

↓ 1 callersFunctioncreate_mixture_csv

Create csv containing mixture information. Each line in the .csv file contains [speech_name, noise_name, noise_onset, noise_offset] Args

prepare_data.py:44

↓ 1 callersFunctionget_amplitude_scaling_factor

Given s and n, return the scaler s according to the snr. Args: s: ndarray, source1. n: ndarray, source2. snr: float, SNR

prepare_data.py:300

↓ 1 callersFunctionget_stats

Calculate stats of PESQ.

evaluate.py:120

↓ 1 callersFunctioninference

Inference all test data, write out recovered wavs to disk. Args: workspace: str, path of workspace. tr_snr: float, training SNR

main_dnn.py:218

↓ 1 callersFunctionpack_features

Load all features, apply log and conver to 3D tensor, write out to .h5 file. Args: workspace: str, path of workspace. data_type

prepare_data.py:369

↓ 1 callersFunctionplot_training_stat

Plot training and testing loss. Args: workspace: str, path of workspace. tr_snr: float, training SNR. bgn_iter: int, pl

evaluate.py:15

↓ 1 callersFunctionreal_to_complex

Recover pred spectrogram's phase from ground truth's phase. Args: pd_abs_x: 2d array, (n_time, n_freq) gt_x: 2d complex array, (

spectrogram_to_wave.py:43

↓ 1 callersFunctionrecover_wav

Recover wave from spectrogram. If you are using scipy.signal.spectrogram, you may need to multipy a scaler to the recovered audio after using

spectrogram_to_wave.py:11

↓ 1 callersFunctiontrain

Train the neural network. Write out model every several iterations. Args: workspace: str, path of workspace. tr_snr: float, tra

main_dnn.py:60

Method__init__

(self, batch_size, type, te_max_iter=None)

data_generator.py:4

Functioninverse_scale_on_2d

Inverse scale 2D array data.

Functionnp_mean_absolute_error

(y_true, y_pred)

prepare_data.py:537

Functionrecover_gt_wav

Recover ground truth wav.

spectrogram_to_wave.py:73