Hybrid Speech Touch Text Entry Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Mobile devices face challenges with accurate and efficient text entry due to small touch screen sizes and susceptibility to errors in noisy environments, particularly when using automatic speech recognition for phone dialing and text input.
Innovation Solution
A system that combines speech and touch input for mobile devices, utilizing a text recognition component, a voice recognition component, and a predictive component to generate a textual output by concatenating observations from both inputs, employing Hidden-Markov Models or Viterbi decoders for improved accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a virtual keypad on a small touch screen is used for text entry, then the device can capture information, but the accuracy of text entry deteriorates due to the inability to tap precise character locations
Solution Approach 1:
The patent combines speech recognition with touch screen input by merging the observation sequences from both input methods. The speech recognizer and text recognizer both produce observations that are concatenated and processed together through a hidden Markov model, allowing the system to leverage both input modalities simultaneously to improve overall text entry accuracy
Solution Approach 2:
The patent introduces a predictive component as an intermediary that combines observations from both speech and touch inputs. This mediator processes the concatenated observation sequences and uses probabilistic models to determine the most likely intended text, resolving ambiguities that neither input method could solve alone
2Productivity
If automatic speech recognition is used for text input, then text entry speed improves, but accuracy deteriorates in noisy environments
Solution Approach 1:
The system merges speech recognition observations with touch screen observations into a unified observation sequence. By processing both input types together through the same hidden Markov model, the system can compensate for speech recognition errors in noisy environments using the more reliable touch input data
Solution Approach 2:
The system uses the combined observations from both input methods to provide feedback to the predictive component, which adjusts its probability calculations based on the concordance or discordance between speech and touch inputs. This feedback mechanism allows the system to identify and correct speech recognition errors
3Quantity of substance
If a touch screen keypad is used for lengthy message entry, then the device can handle significant messages, but the time required for text entry increases and errors occur
Solution Approach 1:
The patent merges speech input with touch input to create a hybrid text entry system. Users can speak portions of messages for rapid input while using touch for corrections or precise character selection, significantly reducing the time required to compose lengthy messages compared to touch-only input
Solution Approach 2:
The system allows users to apply speech recognition partially to portions of a message rather than requiring complete speech or complete touch input. This partial action approach enables users to optimize each segment of the message based on the most efficient input method for that particular content
Data Source
AI summary
The invention relates to systems and or methodologies for enabling combined speech and touch inputs for observation symbol mappings. More particularly, the current innovation leverages the commonality of touch screen display text entry and speech recognition based text entry to increase the speed and accuracy of text entry via mobile devices. Touch screen devices often contain small and closely grouped keypads that can make it difficult for a user to press the intended character, by combining touch screen based text entry with speech recognition based text entry the aforementioned limitation can be overcome efficiently and conveniently.