Microsoft patents voice coding that hides speaker identity for secure calls
Microsoft describes a system that uses machine learning to split voice signals into content and speaker details. It sends only the content to a remote system, stripping away most identifying features of the speaker's voice. A digital watermark is also added to the data before it is transmitted over the network.
In the filing’s own words
A method, computer program product, and computing system for securely transmitting voice signals. A speech signal including a content component and a speaker component of a first voice is received at an encoder. The speaker component of the speech signal is processed, using machine learning, to generate a speaker embedding. The content component of the voice signal is processed, using machine learning and based at least on the speaker embedding, to generate a content embedding having minimized speaker information. The content embedding is transmitted to a decoder for restoring the received speech signal.
What exactly they claim (claim 1)
1. A method comprising: receiving, by an encoder, a speech signal including speaker information of a speaker, the encoder including a trained neural network for encoding speech signals into content embeddings for secure transmission over a communications network to a remote automatic speech recognition (ASR) system; encoding, by the trained neural network, the speech signal into a content embedding to obtain an encoded signal, the content embedding including less of the speaker information than the speech signal; applying a neural watermark to the content embedding prior to transmitting the encoded signal over the communications network; and transmitting the encoded signal over the communications network to the remote ASR system.
Facts
- Company
- Microsoft (Microsoft Technology Licensing, LLC)
- Official title
- SYSTEM AND METHOD FOR SECURELY TRANSMITTING VOICE SIGNALS
- Publication number
- US20260279371A1
- Application number
- 19679574
- Filed
- May 15, 2026
- Published
- Sep 17, 2026
- Inventors
- 3
- Claims
- 20
Classified under
G10L 19/24 Physics & computingG10L 21/007 Physics & computingG10L 15/30 Physics & computingMore from Microsoft in Security & privacy
- Microsoft patents system to call 911 for others during video calls Sep 22, 2026
- Microsoft patents system to block fake code suggested by AI Sep 15, 2026
- Microsoft patents one-way data flow to block network attacks Sep 10, 2026
- Microsoft patents security checks for AI chats using user and model profiles Oct 1, 2026
- Microsoft patents math model to spot ransomware attacks in real time Sep 1, 2026
Summaries are simplified from the filing’s own abstract and first claim. A patent filing is not a product announcement.
Patent data and drawings: U.S. Patent and Trademark Office (USPTO). Nebula is not affiliated with or endorsed by the USPTO.