Motion Adaptive Speech Processing for Voice Destination Entry

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current speech recognition systems face challenges in accurately processing voice destination entries due to the vast number of postal addresses and points of interest, sparse training data, and the need for syntactical constraints, which can lead to decreased recognition accuracy and user experience, especially in dynamic environments like vehicles.

Innovation Solution

A method for motion adaptive speech processing that dynamically estimates a user's motion profile using sensor and non-speech data to interpolate language models, creating a motion adaptive model that enhances speech recognition by focusing on relevant destinations based on the user's location, direction, and behavior, thereby personalizing the search space and improving accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If speech recognition systems process all postal addresses and points of interest, then comprehensive coverage is achieved, but recognition accuracy decreases due to the vast search space

Engineering Contradiction:
Improvecoverage of destinationsVSAvoidrecognition accuracy
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent applies local quality by creating motion-adaptive language models that are specific to the user's local motion context rather than using a single global model for all destinations. The system divides the search space by geographic location and motion state, applying specialized language models tailored to each local context (e.g., models for destinations ahead when moving north versus models for destinations when stationary), thereby maintaining comprehensive coverage while improving recognition accuracy through context-specific processing.

Inventive Principle:
Principle #3Local quality

2Measurement precision

If motion adaptive processing is implemented, then recognition accuracy in dynamic environments improves, but system complexity increases

Engineering Contradiction:
Improverecognition accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent implements dynamics by making the language model selection and interpolation adaptive to the user's motion state. The system dynamically adjusts the language model based on real-time motion parameters (velocity, direction, acceleration) rather than using a static model. This allows the system to automatically adapt to changing environmental conditions without requiring manual intervention, improving recognition accuracy in dynamic vehicle environments while managing complexity through automated adaptation.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent applies parameter changes by varying the language model parameters based on motion characteristics. The system interpolates between different language models using motion-derived weights, where the interpolation parameters are determined by motion state (e.g., stationary vs. moving, direction of travel). This allows the system to smoothly transition between different language models based on motion parameters, improving recognition accuracy while managing complexity through parameter-based adaptation rather than structural changes.

Inventive Principle:
Principle #35Parameter changes

3Productivity

If sparse training data is used for language models, then model training is faster, but recognition accuracy deteriorates

Engineering Contradiction:
Improvemodel training speedVSAvoidrecognition accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies preliminary action by pre-computing and storing multiple location-specific language models during an offline training phase. Instead of training a single comprehensive model with all possible destinations, the system pre-trains specialized language models for different geographic locations and motion contexts using available training data. These pre-trained models are then interpolated in real-time based on the user's motion state, allowing the system to achieve high recognition accuracy without requiring extensive real-time training data processing.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3308379B1Motion adaptive speech processing
Publication Date: 2021.07.28 CERENCE OPERATING CO
  • EP3308379B1 patent drawingFigure 1
  • EP3308379B1 patent drawingFigure 2
  • EP3308379B1 patent drawingFigure 3

AI summary

A method or associated system for motion adaptive speech processing includes dynamically estimating a motion profile that is representative of a user's motion based on data from one or more resources, such as sensors and non-speech resources, associated with the user. The method includes effecting processing of a speech signal received from the user, for example, while the user is in motion, the processing taking into account the estimated motion profile to produce an interpretation of the speech signal. Dynamically estimating the motion profile can include computing a motion weight vector using the data from the one or more resources associated with the user, and can further include interpolating a plurality of models using the motion weight vector to generate a motion adaptive model. The motion adaptive model can be used to enhance voice destination entry for the user and re-used for other users who do not provide motion profiles.