US 12412561

Here's a concise summary of US Patent 12412561, based on the provided authoritative patent text:

US Patent 12412561 Summary

• Title: Real time correction of accent in speech audio signals
• Assignee: Sanas Ai Inc.
• Inventors: Andrei Golman, Dmitrii Sadykov
• Filing Date: 2023-07-31
• Issue Date: 2025-09-09 (Application granted)
• Abstract: Systems and methods for real-time correction of an accent in an input audio signal are provided. A method includes extracting acoustic features from a chunk of a stream of chunks of the input audio signal by an acoustic features extraction module of a computational graph; extracting, by a linguistic features extraction module of the computational graph, linguistic features with a reduced accent from the chunk; synthesizing, by a synthesis module of the computational graph, a spectrum representation based on the acoustic features, the linguistic features, and a speaker embedding for a human speaker; and generating, by a vocoder of the computational graph and based on the spectrum representation, an output chunk of an output audio signal. The input audio signal is digitized with a first sample rate and the output audio signal is digitized with a second sample rate.

Plain-Language Overview of Independent Claims:

The provided patent text does not explicitly list "independent claims" by number, but rather describes a method and system in various embodiments. The "Summary" section, however, outlines the core inventive concepts, which often align with independent claims. Based on the summary and method descriptions (e.g., Method 1100, Method 1300), the key independent aspects relate to:

1. A Method for Real-Time Accent Correction: This method involves several steps performed by a computing system:

   • Extracting acoustic features (e.g., pitch, energy, VAD) from a segment ("chunk") of an incoming audio signal using an acoustic features extraction module within a computational graph.
   • Extracting linguistic features with a reduced accent from the same audio chunk using a linguistic features extraction module, also within the computational graph.
   • Synthesizing a spectrum representation (e.g., melspectrogram with reduced accent) based on the extracted acoustic features, the accent-reduced linguistic features, and a speaker embedding for the speaker, using a synthesis module in the computational graph.
   • Generating an output audio segment ("output chunk") from the spectrum representation using a vocoder in the computational graph.
   • This method is further characterized by the input and output audio signals potentially having different sample rates, with resampling performed between modules as needed. The computational graph can utilize parallel processing units for different modules, and a time-shift parameter can be used to manage delays and synchronize data, where acoustic features may have a lower time-shift than linguistic features.

2. A System for Real-Time Accent Correction: This system includes a computational graph comprising:

   • An acoustic features extraction module configured to extract acoustic features from a chunk of an input audio signal.
   • A linguistic features extraction module configured to extract linguistic features with a reduced accent from the chunk.
   • A synthesis module configured to synthesize a spectrum representation based on the acoustic features, linguistic features, and a speaker embedding.
   • A vocoder configured to generate an output audio chunk based on the spectrum representation.

3. A Non-Transitory Processor-Readable Medium: This medium stores processor-readable instructions that, when executed by a processor, cause the processor to implement the method for real-time accent correction described above.

Litigation Information (as of April 26, 2026):

The patent indicates current litigation:

• First worldwide family litigation filed: This is noted with a link to Darts-ip, but specific details of the litigation (e.g., parties, court, status) are not provided in the accessible text.
• PTAB case PGR2026-00032 filed (Pending): This case is listed as pending, with Unified Patents as the petitioner.

I do not have authoritative information on the specific details of the "First worldwide family litigation" beyond its mention, nor the current status of the PTAB case beyond "Pending" as of the provided patent data fetch date (2026-05-12). No CAFC 2026 dockets were directly returned in the provided patent text, only the PTAB case.