Data

class untwist.data.Signal

Time domain signal. Layout is one column per channel.

samples: ndarray

Signal data.

sample_rate: int

Sample rate in samples / second.

as_ndarray()

Return the data as ndarray again

check_mono()

Utility for ensuring the signal is mono (one channel)

num_channels

Number of channels

num_frames

Number of frames (samples)

class untwist.data.Wave(samples, sample_rate)

Audio waveform signal.

samples: ndarray

Signal data.

sample_rate: int

Sample rate in samples / second.

duration

normalize()

Normalize by maximum amplitude.

play(stop_func=None)

Play the sound with the current audio driver.

stop_func: function

Function to execute when the sound ends.

plot()

Plot the waveofrm using matplotlib.

classmethod read(filename)

Read an audio file (only wav is supported).

filename: string

Path to the wav file.

classmethod record(max_seconds=10, num_channels=2, sr=44100, stop_func=None)

stop()

Stop playback if playing.

write(filename)

Write the data to an audio file (only wav is supported).

filename: string

Path to the wav file.

zero_pad(start_frames, end_frames=0)

Pad with zeros at the start and/or end

start_frames: int

Number of zeros at the start.

end_frames: int

Number of zeros at the end.

class untwist.data.Spectrum

Audio spectrum complex signal.

magnitude()

Return the magnitude spectrum.

phase()

Return the phase spectrum.

plot()

Plot magnitude and phase.

class untwist.data.Spectrogram

Complex audio spectrogram matrix. Rows are frequency bins (0th is the lowest frequency), columns are time bins.

samples: complex

Spectrogram data.

sample_rate: int

Sample rate in samples / second of the original time domain signal.

window_size: int

Window size of the time-frequency transform used to obtain the spectrogram.

hop_size: int

Hop size of the time-frequency transform used to obtain the spectrogram.

magnitude_plot(colormap='CMRmap', min_freq=0, max_freq=None, axes=None, label_x=True, label_y=True, title=None, colorbar=True, log_mag=True)

Plot the magnitude spectrogram

colormap: string

Matplotlib colormap.

min_freq: float

minimum frequency in Hz (for labelling the axis).

max_freq: float

maximum frequency in Hz (for labelling the axis).

axes: matplotlib axes object

Axes object for plotting on existing figure.

label_x: boolean

Add labels to x axis.

label_y: boolean

Add labels to y axis.

title: string

Plot title (overlaid on image).

colorbar: boolean

Add a colorbar.

log_mag: boolean

Plot log magnitude.

num_channels

num_frames

Number of spectral frames.

plot(**kwargs)

class untwist.data.BinaryMask

Binary Mask based on a comparison between target and background. If the threshold is 0, the mask is 1 when the target magnitude is larger than the background, and 0 otherwise.

class untwist.data.RatioMask

Ratio Mask: soft mask based on ratio of target to background magnitude, with optional exponent p.

class untwist.data.Dataset(x_width, x_type, y_width=0, y_type=<type 'numpy.int64'>)

In-memory Dataset

add(x, y=array([], shape=(0, 0), dtype=float64))

get_batch(index, size)

load(path)

normalize()

normalize_points(x)

save(path)

shuffle()

standardize()

standardize_points(x)

class untwist.data.MMDataset(path, x_width=0, x_type=<type 'float'>, y_width=0, y_type=<type 'numpy.int64'>)

Disk-based version using numpy memmap

add(x, y=None)

get_batch(index, size, normalization=2)

load(path)

normalize_points(x)

save()

shuffle()

standardize_points(x)