User-Specific Voice Activity Cutoff Timing for Sentence Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing speech-based human-machine interfaces struggle with determining the end of a sentence, either cutting off slow speakers prematurely or being slow to respond at the end of sentences due to inappropriate timeout periods in voice activity detection.

Innovation Solution

Adaptive adjustment of the voice activity detection (NVAD) cut-off period based on user-specific speech profiles, including typical speech speed and short-term measures, using a continuously adaptive algorithm to dynamically adjust the timeout period.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If a short timeout period is used for voice activity detection, then the system responds quickly to sentence completion, but it cuts off slow speakers or speakers with long pauses prematurely

Engineering Contradiction:

Improveresponse speedVSAvoidsentence completion detection accuracy

Core Design Contradiction:

SpeedVSReliability

Solution Approach 1:

The patent applies dynamics by making the timeout period adjustable rather than fixed. The system dynamically adapts the timeout duration based on detected speech patterns, transitioning from a static short timeout to a flexible timeout that can extend when speech activity is detected, thus resolving the contradiction between quick response and accurate detection

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system changes the timeout parameter dynamically based on speech activity detection. When voice activity is detected during the timeout period, the system extends the timeout duration, effectively changing the parameter from a fixed short value to an adaptive value that matches the speaker's pace, thereby avoiding premature cut-offs while maintaining responsiveness

Inventive Principle:

Principle #35Parameter changes

2Reliability

If a long timeout period is used for voice activity detection, then the system avoids cutting off slow speakers, but it becomes slow to respond at the end of sentences

Engineering Contradiction:

Improvesentence completion detection accuracyVSAvoidresponse time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system dynamically adjusts the timeout period based on real-time speech activity detection. When no speech activity is detected for a certain duration, the system uses a shorter timeout for quick response. When speech activity is detected, the timeout is extended to accommodate slow speakers, thus resolving the contradiction between accuracy and response time

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system uses feedback from voice activity detection to adjust the timeout parameter. The detected speech activity serves as feedback that triggers timeout extension, creating a closed-loop system that adapts to speaker pace and resolves the contradiction between avoiding premature cut-offs and maintaining quick response

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12380889B2Adapting an utterance cut-off period with user specific profile data

Publication Date: 2025.08.05 SOUNDHOUND AI IP LLC

AI summary

A system detects a period of non-voice activity and compares its duration to a cutoff period. The system adapts the cutoff period based on parsing previously-recognized speech of a user that is stored on a user's device or the system, which detects the voice activity, to determine according to a model, such as a machine-learned model, the probability that the speech recognized so far is a prefix to a longer complete utterance. The cutoff period is longer when a parse of previously recognized speech, which is based on the user profile, has a high probability of being a prefix of a longer utterance.