STFT Echo Muter for Speech Recognition Double-Talk

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Acoustic echo from synthesized playback audio in speech-enabled devices interferes with speech recognition, making it difficult for systems to distinguish between user speech and echo, thereby compromising recognition accuracy.

Innovation Solution

A computer-implemented method using an STFT-based echo muter that processes microphone signals to cancel acoustic echo and determine whether frames contain double-talk or echo-only signals, muting echo-only frames to improve speech recognition.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Object-affected harmful factors

If acoustic echo cancellation is applied to remove echo from microphone signals, then echo interference is reduced, but speech recognition accuracy deteriorates when user speech and echo occur simultaneously (double-talk)

Engineering Contradiction:
Improveecho interferenceVSAvoidspeech recognition accuracy
Core Design Contradiction:
Object-affected harmful factorsVSMeasurement precision

Solution Approach 1:

The system dynamically adjusts the echo cancellation strategy based on real-time detection of double-talk conditions. When double-talk is detected (user speech and echo occurring simultaneously), the system switches from aggressive echo cancellation to a mode that preserves speech signals, thereby adapting the processing behavior to current acoustic conditions and resolving the contradiction between echo removal and speech recognition accuracy

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

A double-talk detection mechanism serves as an intermediary between the echo cancellation process and speech recognition system. This intermediary detects whether user speech and echo are occurring simultaneously and controls the echo muting behavior accordingly, preventing the echo canceller from mistakenly muting speech signals while still effectively removing echo interference

Inventive Principle:
Principle #24Intermediary (Mediator)

2Object-affected harmful factors

If echo-only frames are muted to eliminate residual echo, then echo reduction is improved, but speech signals may be inadvertently muted during double-talk periods

Engineering Contradiction:
Improveresidual echoVSAvoidspeech signal preservation
Core Design Contradiction:
Object-affected harmful factorsVSReliability

Solution Approach 1:

The system performs preliminary double-talk detection before applying echo muting operations. By detecting the presence of user speech in advance (before the muting decision is made), the system can prevent inadvertent muting of speech signals while still effectively muting echo-only frames, thus resolving the contradiction between echo reduction and speech signal preservation

Inventive Principle:
Principle #10Preliminary action

3Reliability

If Double-talk Detection is implemented to distinguish between double-talk and echo-only frames, then speech recognition reliability is improved, but system complexity increases

Engineering Contradiction:
Improvespeech recognition reliabilityVSAvoidsignal processing complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The double-talk detection functionality is extracted as a separate, dedicated module that operates independently from the echo cancellation and speech recognition components. This modular extraction allows the system to achieve improved speech recognition reliability through accurate double-talk detection while managing complexity by isolating the detection logic in a self-contained unit with clear interfaces to other system components

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP4385011B1STFT-based echo muter
Publication Date: 2025.06.18 GOOGLE LLC
  • EP4385011B1 patent drawingFigure 1
  • EP4385011B1 patent drawingFigure 2
  • EP4385011B1 patent drawingFigure 3

AI summary

A method (300) for Short-Time Fourier Transform-based echo muting includes receiving a microphone signal (202) including acoustic echo (156) captured by a microphone and corresponding to audio content (154) from an acoustic speaker (118), and receiving a reference signal (158) including a sequence of frames representing the audio content. For each frame, the method includes processing, using an acoustic echo cancel er (210) configured to receive a respective frame as input to generate a respective output signal frame (206) that cancels the acoustic echo from the respective frame, and determining, using a Double-talk Detector (220), based on the respective frame and output signal frame, whether the respective frame includes a double-talk frame or an echo-only frame. For each respective frame that includes the echo-only frame, muting the respective output signal frame, and performing speech processing on the respective output signal frame for each respective frame that includes the double-talk frame.