emonet.batch_vad module

Module to apply pre-train VAD model to files.

Intended to run as batch process before any training begins.

class emonet.batch_vad.VadChunk(model: torch.nn.modules.module.Module, sample_rate: int = 16000)

Bases: object

Class for applying pre-trained VAD model to audio file.

Constructor method.

Parameters

• model (nn.Module) – Pre-trained voiced activity detection (VAD) model.

• sample_rate (int) – Sample rate of audio file.

__init__(model: torch.nn.modules.module.Module, sample_rate: int = 16000)

Constructor method.

Parameters

• model (nn.Module) – Pre-trained voiced activity detection (VAD) model.

• sample_rate (int) – Sample rate of audio file.

get(x: torch.Tensor, filename: pathlib.Path) → torch.Tensor

Get speech chunks consolidated into single tensor.

Parameters

• x (torch.Tensor) – Original audio signal.

• filename (pathlib.Path) – Filename containing original audio for tensor x.

Returns

torch.Tensor – VAD chunks consolidated into single tensor.

emonet.batch_vad.main()

Run VAD for all wav files in data directory.