Context-Aware Speech Recognition for Mobile Call Keyword Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Users face inconvenience in accessing context information during phone conversations, as they need to manually open applications on their mobile devices despite the information being available, due to limitations in speech recognition accuracy for non-user voices and the need for user-specific training data.

Innovation Solution

A method and apparatus that establish a call connection between terminals to monitor conversations for predetermined context-related keywords, providing an indication to users to open linked applications, such as calendar, contact, or map applications, by utilizing speech recognition and context data sharing between devices.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If speech recognition application requires user-specific training data to improve accuracy, then recognition accuracy for registered users improves, but the system becomes less accessible to non-users and requires additional setup time

Engineering Contradiction:

Improvespeech recognition accuracyVSAvoidaccessibility to non-users

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent uses contact information from the device's contact list as a template or copy to automatically populate speech recognition training data. Instead of requiring users to manually provide training samples, the system copies relevant information (name, phone number, associated context) from existing contact records to create training datasets, thereby maintaining high accuracy while eliminating the need for users to perform manual training setup

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The system performs automatic speech recognition training in the background without requiring user intervention. The device self-services by automatically detecting contact information, extracting relevant features, and using them to train the speech recognition model, thus providing both high accuracy and immediate accessibility to non-users without manual setup

Inventive Principle:

Principle #25Self-service

2Measurement precision

If user manually opens applications to check context information during phone conversation, then information accuracy is ensured, but usability and convenience deteriorate

Engineering Contradiction:

Improveinformation accuracyVSAvoidusability during call

Core Design Contradiction:

Measurement precisionVSEase of operation

Solution Approach 1:

The patent replaces the mechanical interaction of manually opening applications with an automated speech-based system. The speech recognition application listens to the conversation, automatically detects context-related keywords, and triggers the appropriate applications without requiring physical manipulation of the device or manual navigation through menus during the call

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The speech recognition application acts as an intermediary between the phone conversation and the context information applications. It monitors the audio stream, interprets spoken keywords, and mediates by automatically launching the relevant applications (calendar, contacts, maps) based on the detected context, thus maintaining information accuracy while eliminating manual operations

Inventive Principle:

Principle #24Intermediary (Mediator)

3Productivity

If speech recognition application monitors conversation continuously to detect keywords, then context information retrieval improves, but energy consumption and processing load increase

Engineering Contradiction:

Improvecontext information retrieval speedVSAvoidenergy consumption during call

Core Design Contradiction:

ProductivityVSUse of energy by moving object

Solution Approach 1:

The system extracts and processes only specific context-related keywords from the continuous conversation stream rather than analyzing every word. By pre-defining a set of relevant keywords (such as meeting-related terms, contact names, location references) and selectively monitoring for these specific terms, the system achieves fast context retrieval while significantly reducing the processing load and energy consumption compared to full conversation analysis

Inventive Principle:

Principle #2Taking out (Extraction)

Data Source

PatentUS9224396B2Enhanced context awareness for speech recognition

Publication Date: 2015.12.29 PIECE FUTURE PTE LTD

AI summary

A method comprising establishing a call connection (200) between at least a first and a second terminal; monitoring (202), by at least the first terminal, a conversation during the call in order to detect (204) at least one predetermined context-related keyword repeated in at least the first and the second terminal; and in response to detecting (210) at least one repeated predetermined context-related keyword, providing an indication (212, 214, 216) about the detected context-related keyword to a user of at least the first terminal, said indication enabling opening an application linked to said context-related keyword.