Encrypted Word Sequence Adaptation for Speech Recognition Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech recognition systems for hard-of-hearing individuals face challenges with inaccurate and delayed transcriptions, which hinder effective participation in audio communications.
Innovation Solution
Implementing a transcription system that combines multiple automatic speech recognition (ASR) systems and revoicing systems to generate a single, fused transcription, along with real-time accuracy estimation and correction mechanisms to enhance transcription accuracy and reduce latency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple ASR systems are used to improve transcription accuracy, then transcription accuracy is improved, but device complexity increases
Solution Approach 1:
The patent combines multiple ASR systems into a unified transcription system that processes audio data through several parallel ASR engines. Each ASR system generates transcriptions that are then fused together using a combination mechanism that weights and integrates results from different systems, thereby improving overall transcription accuracy while managing system complexity through structured integration.
Solution Approach 2:
The patent introduces a transcription fusion component as an intermediary between multiple ASR systems and the final output. This mediator receives transcriptions from multiple ASR systems, applies weighting algorithms, and combines them into a single fused transcription, thereby managing the complexity of coordinating multiple systems while achieving improved accuracy through diverse input sources.
2Loss of time
If real-time transcription processing is implemented, then latency is reduced, but computational load increases
Solution Approach 1:
The patent implements preliminary processing steps where audio data is pre-prepared and pre-processed before being fed into multiple ASR systems. Transcriptions from different ASR systems are prepared in parallel, and the fusion process uses pre-computed weighted scores to quickly combine results, thereby reducing real-time processing latency while managing computational load through advance preparation.
Solution Approach 2:
The patent applies partial processing by using confidence thresholds to determine when full multi-ASR fusion is necessary versus when a single ASR output suffices. When transcription confidence is high, the system accepts partial results without full fusion, reducing computational load. When confidence is low or accuracy is critical, the system performs excessive action by fully processing all ASR outputs through the fusion mechanism.
3Measurement precision
If transcription fusion with multiple ASR systems is used, then transcription accuracy is improved, but processing time increases
Solution Approach 1:
The patent implements periodic updates of transcription results where the fusion process operates in intervals rather than continuously. Audio data is processed in chunks, and transcriptions are fused at periodic intervals, allowing the system to maintain accuracy through multiple ASR inputs while reducing processing time by avoiding continuous full-fusion operations. This periodic approach balances accuracy requirements with processing efficiency.
Data Source
AI summary
A method may include obtaining a text string that is a transcription of audio data and selecting a sequence of words from the text string as a first word sequence. The method may further include encrypting the first word sequence and comparing the encrypted first word sequence to multiple encrypted word sequences. Each of the multiple encrypted word sequences may be associated with a corresponding one of multiple counters. The method may also include in response to the encrypted first word sequence corresponding to one of the multiple encrypted word sequences based on the comparison, incrementing a counter of the multiple counters associated with the one of the multiple encrypted word sequences and adapting a language model of an automatic transcription system using the multiple encrypted word sequences and the multiple counters.