NeMo
NeMo
speech
audio

SE Denoising SB 16kHz Small

Model architecture | Model size

Model Overview

Description

The model extracts speech for human or machine listeners. This is a generative speech denoising model based on the Schrödinger bridge. The model is trained on a publicly available research dataset.

This model is for research and development only.

License/Terms of Use

License to use this model is covered by the CC-BY-NC-SA-4.0. By downloading the public and release version of the model, you accept the terms and conditions of the CC-BY-NC-SA-4.0 license.

References

[1] Schrödinger Bridge for Generative Speech Enhancement, Interspeech, 2024.

Model Architecture

Architecture Type: Schrödinger Bridge
Network Architecture: U-Net with convolutional layers

Input

Input Type(s): Audio
Input Format(s): .wav files
Input Parameters: One-Dimensional (1D)
Other Properties Related to Input: 16000 Hz Mono-channel Audio

Output

Output Type(s): Audio
Output Format: .wav files
Output Parameters: One-Dimensional (1D)
Other Properties Related to Output: 16000 Hz Mono-channel Audio

Software Integration

Runtime Engine(s):

  • NeMo-2.0.0

Supported Hardware Microarchitecture Compatibility:

  • NVIDIA Ampere
  • NVIDIA Blackwell
  • NVIDIA Jetson
  • NVIDIA Hopper
  • NVIDIA Lovelace
  • NVIDIA Turing
  • NVIDIA Volta

Preferred Operating System(s)

  • Linux
  • Windows

Model Version(s)

se_den_sb_16k_small_v1.0

Training, Testing, and Evaluation Datasets

Training Dataset

Link: WSJ0, CHiME3

Data Collection Method by dataset: Human

Labeling Method by dataset: Human

Properties (Quantity, Dataset Descriptions, Sensor(s)): WSJ0 was used for clean speech signals and CHiME3 was used for additive noise signals. The observed signals are generated with signal-to-noise ratios between -6dB and 14dB. The total size of the training dataset was approximately 25 hours.

Testing Dataset

Link: WSJ0, CHiME3

Data Collection Method by dataset: Human

Labeling Method by dataset: Human

Properties (Quantity, Dataset Descriptions, Sensor(s)): WSJ0 was used for clean speech signals and CHiME3 was used for additive noise signals. The observed signals are generated with signal-to-noise ratios between -6dB and 14dB. The total size of the testing dataset was approximately 2 hours.

Evaluation Dataset

Link: WSJ0, CHiME3

Data Collection Method by dataset: Human

Labeling Method by dataset: Human

Properties (Quantity, Dataset Descriptions, Sensor(s)): WSJ0 was used for clean speech signals and CHiME3 was used for additive noise signals. The observed signals are generated with signal-to-noise ratios between -6dB and 14dB. The total size of the evaluation dataset was approximately 2 hours.

Inference

Engine: NeMo 2.0

Test Hardware: NVIDIA V100

Performance

The model is trained on the training subset of the WSJ0-CHiME3 dataset using the auxiliary L1-norm loss [1].

The model is evaluated using several instrumental metrics: perceptual evaluation of speech quality (PESQ), extended short-term objective intelligibility (ESTOI) and scale-invariant signal-to-distortion ratio (SI-SDR). Word error rate (WER) is evaluated using the FastConformer-Transducer-Large English ASR model.

Metrics are reported on the test set of WSJ0-CHiME dataset using either SDE or ODE sampler.

Signal PESQ ESTOI SI-SDR/dB WER / %
Input 1.35 0.63 4.0 12.18
Processed SDE 2.67 0.89 15.1 5.10
Processed ODE 2.77 0.90 16.2 4.13

How to use this model

The model is available for use in the NVIDIA NeMo toolkit, and can be used to process audio or for fine-tuning.

Load the model

from nemo.collections.audio.models import AudioToAudioModel
model = AudioToAudioModel.from_pretrained('nvidia/se_den_sb_16k_small')

Process audio

A single audio file can be processed as follows

import librosa
audio_in, _ = librosa.load(path_to_input_audio, sr=model.sample_rate)
audio_in_signal = torch.from_numpy(audio_in).view(1, 1, -1).to(device)
audio_in_length = torch.tensor([audio_in_signal.size(-1)]).to(device)

audio_out_signal, _ = model(input_signal=audio_in_signal, input_length=audio_in_length)

For processing several audio files at once, check the process_audio script in NeMo.

Listen to audio

import soundfile as sf
audio_out = audio_out_signal.cpu().numpy().squeeze()
sf.write(path_to_output_audio, audio_out, samplerate=model.sample_rate)

Change sampler configuration

model.sampler.process = 'ode' # default sampler is 'sde'
model.sampler.num_steps = 10 # default is 50 steps

audio_out_signal, _ = model(input_signal=audio_in_signal, input_length=audio_in_length)

Ethical Considerations

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.

Please report security vulnerabilities or NVIDIA AI Concerns here.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference API
Unable to determine this model’s pipeline type. Check the docs .