Multi-Channel Speech Compression Using Visual Source Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multi-channel speech processing systems face limitations due to insufficient bandwidth for transmitting raw audio signals from multiple recording devices, necessitating more processing by back-end systems and reducing efficiency.
Innovation Solution
The system employs a method to encode audio encounter information using a reference audio acquisition device, estimates location information via machine vision, and selects acoustic relative transfer functions based on this information to enhance coding and compression, leveraging the fixed geometry of audio recording devices for improved speech signal processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multi-channel audio signals are transmitted from front-end recording devices to back-end processing systems, then speech processing capability is improved, but bandwidth requirements increase beyond available capacity
Solution Approach 1:
The patent extracts and transmits only the essential acoustic transfer function information between reference and non-reference channels to the back-end system, rather than transmitting complete multi-channel audio signals. This extraction approach reduces bandwidth requirements while preserving the critical spatial and acoustic characteristics needed for effective speech processing.
Solution Approach 2:
The system creates a compressed representation (copy) of the acoustic transfer functions that captures the essential relationships between multiple audio channels. This copy contains sufficient information for the back-end system to reconstruct or synthesize the necessary audio processing capabilities without requiring the full original multi-channel signal data.
2Measurement precision
If raw audio signals from multiple recording devices are transmitted to back-end system, then speech processing accuracy is improved, but processing load on back-end system increases
Solution Approach 1:
The front-end recording system performs preliminary processing by estimating and encoding the acoustic transfer functions that characterize the acoustic relationships between different microphones. This preliminary action transforms the raw multi-channel audio data into a more compact representation that preserves speech processing accuracy while reducing the computational burden on the back-end system.
Solution Approach 2:
The system changes the representation parameters from raw time-domain audio waveforms to frequency-domain acoustic transfer function coefficients. This parameter transformation maintains the essential speech information needed for accurate processing while significantly reducing the amount of data that requires back-end processing power.
3Reliability
If all raw audio channels are transmitted to back-end system, then audio quality is maintained, but transmission bandwidth becomes insufficient
Solution Approach 1:
The patent applies different processing and transmission strategies to different audio channels based on their relative importance and acoustic characteristics. By identifying a reference channel and calculating transfer functions from it to other channels, the system maintains high audio quality for the reference channel while using compressed transfer function representations for other channels, thereby optimizing the quality-bandwidth trade-off.
Data Source
AI summary
A method, computer program product, and computing system for encoding audio encounter information of a reference audio acquisition device of a plurality of audio acquisition devices of an audio recording system, thus defining encoded reference audio encounter information. Location information may be estimated, via a machine vision system, for an acoustic source within an acoustic environment. One or more acoustic relative transfer functions may be selected from a plurality of acoustic relative transfer functions for the plurality of audio acquisition devices of the audio recording system based upon, at least in part, the location information. The encoded reference audio encounter information and a representation of the selected one or more acoustic relative transfer function may be transmitted.