18416589. ELECTRONIC DEVICE AND METHOD OF LOW LATENCY SPEECH ENHANCEMENT USING AUTOREGRESSIVE CONDITIONING-BASED NEURAL NETWORK MODEL simplified abstract (Samsung Electronics Co., Ltd.)

ELECTRONIC DEVICE AND METHOD OF LOW LATENCY SPEECH ENHANCEMENT USING AUTOREGRESSIVE CONDITIONING-BASED NEURAL NETWORK MODEL

Organization Name

Samsung Electronics Co., Ltd.

Inventor(s)

Nikolas Andrew Babaev of Moscow (RU)

Pavel Konstantinovich Andreev of Moscow (RU)

Azat Rustamovich Saginbaev of Moscow (RU)

Ivan Sergeevich Shchekotov of Moscow (RU)

ELECTRONIC DEVICE AND METHOD OF LOW LATENCY SPEECH ENHANCEMENT USING AUTOREGRESSIVE CONDITIONING-BASED NEURAL NETWORK MODEL - A simplified explanation of the abstract

This abstract first appeared for US patent application 18416589 titled 'ELECTRONIC DEVICE AND METHOD OF LOW LATENCY SPEECH ENHANCEMENT USING AUTOREGRESSIVE CONDITIONING-BASED NEURAL NETWORK MODEL

Simplified Explanation

A neural method model is trained using teacher forcing and autoregressive channels, with ground-truth shifted waveforms and predictions being used in training iterations.

• Neural network model trained using teacher forcing and autoregressive channels
• Ground-truth shifted waveform used in initial training iteration
• Predictions of neural network model used in subsequent training iterations
• Inference performed using additional channel containing predictions
• Speech enhancement performed using trained neural network model

Potential Applications

The technology described in this patent application could be applied in the following areas:

• Speech enhancement systems
• Audio processing applications
• Noise reduction in audio signals

Problems Solved

This technology addresses the following issues:

• Improving speech quality in noisy environments
• Enhancing audio signals for better clarity
• Reducing background noise in audio recordings

Benefits

The benefits of this technology include:

• Enhanced speech intelligibility
• Improved audio quality
• Better performance in noisy conditions

Potential Commercial Applications

The technology could be utilized in various commercial applications such as:

• Audio editing software
• Communication devices
• Voice recognition systems

Possible Prior Art

One possible prior art for this technology could be the use of autoregressive models in speech enhancement systems.

How does this technology compare to existing speech enhancement methods?

This article does not provide a direct comparison to other speech enhancement methods currently available in the market.

What is the computational complexity of the neural method model described in the patent application?

The article does not delve into the computational complexity of the neural method model and how it compares to other similar models.

Original Abstract Submitted

A neural method model is trained by, in an initial training iteration, training the neural network model in a teacher forcing mode in which an autoregressive channel includes a ground-truth shifted waveform, and outputting predictions of the neural network model; and in at least one additional training iteration, replacing the ground-truth shifted waveform in the autoregressive channel with the predictions of the neural network model obtained in a previous training iteration. An inference may then be performed by providing, for the neural network model, an additional channel containing at least one prediction of the neural network model outputted during training; and performing speech enhancement using the neural network model.