ASR Audio Stream Compaction for Lower Transcription Charges
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio stream processing systems incur high costs due to fixed pricing models for speech-to-text services, regardless of audio file duration, leading to inefficiencies and increased expenses for enterprises handling large volumes of short audio streams.
Innovation Solution
Implementing transaction-based and time-based audio stream processing techniques, where audio streams are optimized before transmission to ASR services, either by concatenating multiple streams into a single payload or compacting them to remove non-meaningful content, based on factors like network speed and cost structures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If audio streams are transmitted individually to ASR services, then transcription accuracy is maintained, but transcription costs increase due to fixed pricing models
Solution Approach 1:
The patent combines multiple short audio streams into a single concatenated audio stream for batch processing by ASR services. This merging approach reduces the number of individual transcription requests, thereby lowering costs under fixed pricing models while maintaining transcription accuracy through proper stream separation markers
Solution Approach 2:
The system performs preliminary actions by pre-processing audio streams before transmission to ASR services. This includes compacting audio data, removing silence periods, and organizing streams into batches, which optimizes transmission efficiency and reduces transcription costs before the actual transcription process occurs
2Productivity
If audio streams are compacted to remove non-meaningful content, then transmission frequency increases and costs decrease, but processing complexity increases
Solution Approach 1:
The patent extracts and removes non-meaningful content from audio streams, such as silence periods and redundant segments, before transmission. This extraction process reduces audio file sizes and increases transmission frequency while managing processing complexity through targeted removal of specific content types
3Loss of energy
If multiple audio streams are concatenated into a single payload, then ASR service costs are reduced, but stream separation and identification become more difficult
Solution Approach 1:
The patent introduces intermediary separation markers or delimiters between concatenated audio streams. These markers serve as mediators that enable the ASR service to easily identify and separate individual streams within the combined payload, reducing costs while eliminating the difficulty of stream separation through clear structural boundaries
4Reliability
If transcription services are used for every audio stream regardless of duration, then service coverage is complete, but cost efficiency decreases for short audio files
Solution Approach 1:
The patent applies partial action by selectively processing audio streams based on their characteristics. Instead of transcribing every audio stream individually, the system batches multiple short streams together for combined transcription, achieving complete service coverage while improving cost efficiency through optimized resource utilization
Data Source
AI summary
Systems and methods for processing audio streams are disclosed herein. An audio stream including speech content is received. The audio stream is compacted to generate a compacted audio stream and the compacted audio stream is transmitted to an automatic speech recognition (ASR) service for transcription of the speech content to text content. In response to transmitting the compacted audio stream for transcription, text content, a transcription of the audio stream, is received from the ASR service.