Visual Targeting Alignment for Time-Sensitive Voice Commands

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing systems face delays and inaccuracies in processing voice commands due to the time-based nature of human speech and rapid changes in visual targeting, leading to misapplication of commands to unintended objects.

Innovation Solution

Implementing proactive, reactive, and real-time targeting analysis techniques to identify visual targets before, during, or after spoken utterances, utilizing gaze and speech recognition to align audio and visual data streams for accurate command association.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If voice commands are processed sequentially after speech recognition, then speech recognition accuracy is maintained, but command association with visual targets becomes inaccurate due to time delays

Engineering Contradiction:
Improvespeech recognition accuracyVSAvoidcommand association accuracy
Core Design Contradiction:
Measurement precisionVSManufacturing precision

Solution Approach 1:

The system performs preliminary action by capturing visual targeting data (gaze direction, head orientation) in advance and storing it in a buffer before the voice command is fully processed. This allows the visual target to be pre-identified and associated with the upcoming voice command, eliminating the time delay problem where the target might have changed by the time processing completes.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent introduces an intermediary mechanism - a temporal alignment system that uses timestamps and buffered data to bridge the gap between visual targeting detection and voice command processing. This intermediary layer synchronizes the two data streams (visual and audio) in time, ensuring accurate association even when processing occurs sequentially.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Speed

If visual targeting data is captured in real-time, then responsiveness to user focus changes is improved, but processing latency increases due to continuous data collection and alignment

Engineering Contradiction:
Improveresponse to focus changesVSAvoidprocessing latency
Core Design Contradiction:
SpeedVSLoss of time

Solution Approach 1:

The system implements periodic action by capturing visual targeting data at discrete time intervals and buffering it, rather than continuously processing it in real-time. This periodic capture approach maintains responsiveness to focus changes while reducing processing latency by batch-processing the captured data when needed, rather than requiring immediate continuous alignment.

Inventive Principle:
Principle #19Periodic action

Solution Approach 2:

Visual targeting data is captured and buffered in advance (preliminary action) before the voice command processing needs it. This allows the system to prepare the data ahead of time, reducing the processing latency when the voice command is detected, while still maintaining real-time responsiveness through continuous buffering.

Inventive Principle:
Principle #10Preliminary action

3Reliability

If the system waits for complete speech recognition before identifying visual targets, then speech processing is thorough, but commands may be misapplied to unintended objects due to rapid user focus changes

Engineering Contradiction:
Improvespeech processing thoroughnessVSAvoidtarget identification accuracy
Core Design Contradiction:
ReliabilityVSManufacturing precision

Solution Approach 1:

The system performs preliminary action by capturing and buffering visual targeting data before the voice command is fully processed. This allows the visual target to be identified and locked in advance, ensuring that even if speech processing takes time, the visual target remains associated with the correct object and not shifted due to rapid user focus changes during the processing delay.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3909044B1Time-based visual targeting for voice commands
Publication Date: 2026.03.25 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP3909044B1 patent drawingFigure 1
  • EP3909044B1 patent drawingFigure 2
  • EP3909044B1 patent drawingFigure 3

AI summary

A method performed by a computing system for directing a voice command to a function associated with a visual target includes receiving a set of time-variable sensor-based data streams, including an audio data stream and a targeting data stream. The targeting data stream is stored in a buffer as buffered targeting data. Presence of a spoken utterance is identified within the audio data stream and is associated with a temporal identifier corresponding in time to the set of sensor-based data streams. A voice command corresponding to the spoken utterance is identified. A visual targeting vector within the buffered targeting data and a visual target of that visual targeting vector is identified at a time corresponding to the temporal identifier. The voice command is directed to a function associated with the visual target to generate an output.