Trusted Speech Transcription via Digital Signature Linking
Solution Overview
Problem
Current speech recording technologies face challenges in linking transcribed speech to the original voice record, making it difficult to identify speakers reliably, and require excessive storage space, which hinders legal and business applications relying on voice identification.
Innovation Solution
A system that digitally signs and links transcribed text with corresponding voice recordings using a trusted execution environment (TEE), incorporating audio sample selection, speech-to-text transcription, and packaging elements to create a non-repudiable transcription pair, ensuring the authenticity and integrity of the speech data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If speech is transcribed to text format, then storage space is reduced and searchability is improved, but the ability to identify speakers reliably is lost
Solution Approach 1:
The system segments the speech data into two distinct components: (1) full speech recordings stored in raw audio formats for speaker identification, and (2) transcribed text segments stored in text formats for search and processing. This segmentation allows each component to serve its specific function optimally without compromising the other.
Solution Approach 2:
The patent implements a nested structure where transcribed text is embedded within a container that also holds references to the original speech segments. The container format nests multiple transcription segments and their corresponding speech references together, creating a hierarchical structure that preserves the link between text and audio while enabling efficient storage and retrieval.
2Quantity of substance
If only transcribed text is stored, then storage efficiency is improved, but legal and business applications requiring voice identification cannot be supported
Solution Approach 1:
The container format is designed to be multi-functional, serving both storage and verification purposes. It simultaneously stores transcribed text for efficient searching, references to original speech for speaker identification, and digital signatures for authenticity verification. This universal design enables the system to support diverse applications including legal compliance, business analytics, and speaker verification.
Solution Approach 2:
The patent introduces a container format as an intermediary structure that bridges the gap between transcribed text and original speech recordings. This intermediary holds both the text and metadata references to the audio segments, enabling the system to provide speaker identification capabilities without requiring storage of all original audio data, thus maintaining storage efficiency while supporting legal and business applications.
3Reliability
If digital signing is implemented to ensure authenticity, then data integrity is improved, but processing complexity increases
Solution Approach 1:
The system performs digital signing as a preliminary action during the transcription process itself, rather than as a separate post-processing step. The container format is created with embedded digital signatures at the time of transcription, ensuring data integrity from the outset. This preliminary action approach simplifies subsequent processing by eliminating the need for separate verification steps.
Data Source
AI summary
Embodiments include a system, method, and apparatus for creating a trusted speech transcription. Transcription logic can receive a signal of audible speech from an audio source and convert the audible speech signal into text. Sampling logic can receive the signal of the audible speech and record a portion of the of the audible speech as a digital audio recording of the audible speech. Packaging logic can create a trusted speech transcription data record that includes the text and the digital audio recording and to digitally sign the transcription file to securely associate the text with the digital audio recording.