Self-Trained Speech Transcription Engine for Parallel Audio Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current state-of-the-art machine transcription systems convert speech to text at a 1-to-1 ratio of audio file length versus transcribing time, which is inefficient and does not account for real-time processing or accuracy enhancements.
Innovation Solution
An adaptive self-trained computer engine utilizing a multi-core processor dynamically segments and transcribes speech audio data, applying filters and speaker profiles to achieve faster transcription rates and improve accuracy, particularly in activity areas like financial trading.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If current state-of-the-art machine transcription is used, then speech can be converted to text, but the transcription time equals the audio file length (1-to-1 ratio), which is inefficient
Solution Approach 1:
The patent segments the audio file into multiple chunks that can be processed in parallel. Each chunk is transcribed independently by separate processing threads, allowing simultaneous transcription of multiple segments. This parallel processing approach reduces the total transcription time from being proportional to the full audio length to a fraction thereof, directly resolving the contradiction between maintaining transcription quality and reducing transcription time.
2Productivity
If faster transcription is achieved through parallel processing, then productivity increases, but system complexity increases due to multi-core processor requirements
Solution Approach 1:
The patent employs a universal processing framework that can operate on any multi-core processor architecture. The transcription engine is designed to be architecture-agnostic, utilizing standard parallel processing interfaces and memory management mechanisms that are universally supported across different multi-core systems. This allows the system to leverage available hardware parallelism without being tied to specific processor implementations, thereby increasing productivity while minimizing the impact of added complexity.
3Speed
If real-time processing is implemented, then transcription speed exceeds human speech rate, but accuracy may be compromised
Solution Approach 1:
The patent performs preliminary processing actions on audio chunks before full transcription, including preprocessing steps such as noise filtering, feature extraction, and preliminary phoneme identification. These preliminary actions prepare the data in advance, enabling faster subsequent transcription while maintaining accuracy. By doing preparatory work upfront on segmented chunks, the system achieves real-time processing speeds without sacrificing transcription quality.
Solution Approach 2:
The system incorporates feedback mechanisms where transcription results are continuously validated and refined. As each audio chunk is transcribed, the results are checked for accuracy through multiple verification passes, and corrections are applied iteratively. This feedback loop ensures that even as processing speed increases through parallel operations, transcription accuracy is maintained by continuously monitoring and correcting errors across all parallel processing streams.
Data Source
AI summary
In some embodiments, the present invention provides for an exemplary computer system which includes at least the following components: an adaptive self-trained computer engine programmed, during a training stage, to electronically receive an initial speech audio data generated by a microphone of a computing device; dynamically segment the initial speech audio data and the corresponding initial text into a plurality of user phonemes; dynamically associate a plurality of first timestamps with the plurality of user-specific subject-specific phonemes; and, during a transcription stage, electronically receive to-be-transcribed speech audio data of at least one user; dynamically split the to-be transcribed speech audio data into a plurality of to-be-transcribed speech audio segments; dynamically assigning each timestamped to-be-transcribed speech audio segment to a particular core of the multi-core processor; and dynamically transcribing, in parallel, the plurality of timestamped to-be-transcribed speech audio segments based on the user-specific subject-specific speech training model.