17534396. OBFUSCATING AUDIO SAMPLES FOR HEALTH PRIVACY CONTEXTS simplified abstract (International Business Machines Corporation)

OBFUSCATING AUDIO SAMPLES FOR HEALTH PRIVACY CONTEXTS

Organization Name

International Business Machines Corporation

Inventor(s)

Victor Abayomi Akinwande of Karen (KE)

Celia Cintas of Nairobi (KE)

Komminist Weldemariam of Ottawa (CA)

Aisha Walcott of Nairobi (KE)

OBFUSCATING AUDIO SAMPLES FOR HEALTH PRIVACY CONTEXTS - A simplified explanation of the abstract

This abstract first appeared for US patent application 17534396 titled 'OBFUSCATING AUDIO SAMPLES FOR HEALTH PRIVACY CONTEXTS

Simplified Explanation

The patent application describes a system for detecting and removing bio-markers from audio samples using a combination of supervised learning and generative adversarial networks (GANs). Here are the key points:

• The system includes a supervised discriminator that is trained to identify bio-markers in an audio sample dataset.
• A denoising autoencoder is also trained to learn a latent space that can be used to reconstruct an output audio sample with the same quality as the input sample.
• A conditional auxiliary GAN is trained to generate an output audio sample that is free of bio-markers, while maintaining the fidelity of the input sample.
• The system deploys the trained GAN, discriminator, and autoencoder in an audio processing system.

Potential applications of this technology:

• Audio forensics: The system can be used to detect and remove bio-markers from audio recordings, which can be useful in forensic investigations.
• Privacy protection: By removing bio-markers from audio samples, the system can help protect the privacy of individuals who may have inadvertently recorded sensitive information.
• Noise reduction: The denoising autoencoder component of the system can be used to remove unwanted noise from audio recordings.

Problems solved by this technology:

• Detection of bio-markers: The supervised discriminator is trained to accurately identify bio-markers in audio samples, which can be challenging due to variations in audio quality and background noise.
• Bio-marker removal: The conditional auxiliary GAN is trained to generate audio samples that are free of bio-markers, providing a way to remove sensitive information from recordings.

Benefits of this technology:

• Accuracy: The system combines supervised learning and GANs to achieve accurate detection and removal of bio-markers from audio samples.
• Fidelity preservation: The denoising autoencoder ensures that the output audio samples maintain the same quality as the input samples.
• Versatility: The system can be deployed in various audio processing applications, including forensics, privacy protection, and noise reduction.

Original Abstract Submitted

A supervised discriminator for detecting bio-markers in an audio sample dataset is trained and a denoising autoencoder is trained to learn a latent space that is used to reconstruct an output audio sample with a same fidelity as an input audio sample of the audio sample dataset. A conditional auxiliary generative adversarial network (GAN) trained to generate the output audio sample with the same fidelity as the input audio sample, wherein the output audio sample is void of the bio-markers. The conditional auxiliary generative adversarial network (GAN), the corresponding supervised discriminator, and the corresponding denoising autoencoder are deployed in an audio processing system