Mono-Recording Voice Separation for Contact Center Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Contact centers face challenges in accurately analyzing customer and agent interactions due to the lack of separated voice data in mono recordings, hindering quick and accurate analysis of customer and agent communications.
Innovation Solution
A system and method that record a mono recording of customer-agent interactions, separately record agent voice data, align and subtract it from the mono recording to isolate customer voice data, and convert it to text for linguistic analysis to determine personality types and identify distress events.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If mono recording is used to record customer-agent interactions, then the recording process is simple and compatible with standard telephone networks, but the voice data cannot be separated for accurate analysis
Solution Approach 1:
The patent segments the mixed mono recording into separate customer voice data and agent voice data through signal processing. The system divides the composite audio signal by identifying and isolating distinct voice sources, enabling separate analysis of each party's communication while maintaining the simplicity of mono recording capture.
Solution Approach 2:
The patent introduces an intermediary signal processing system that acts as a mediator between the mono recording and the analysis requirements. This intermediary component performs voice separation, alignment, and extraction operations, allowing the system to maintain simple mono recording compatibility while achieving precise voice data separation for analysis.
2Measurement precision
If separated stereo recording is used for customer and agent voices, then accurate voice separation is achieved, but compatibility with standard telephone networks is lost
Solution Approach 1:
The patent transitions from the traditional two-dimensional approach (separate stereo channels) to a different dimension by processing mono recording data through temporal and spectral analysis. Instead of requiring separate physical channels, the system extracts separated voice data from the single-channel mono recording by analyzing time-synchronized patterns and frequency characteristics, achieving separation accuracy without sacrificing network compatibility.
3Productivity
If voice data analysis is performed on unseparated mono recordings, then the analysis process is quick, but the accuracy of customer and agent communication analysis is hindered
Solution Approach 1:
The patent performs preliminary voice separation and alignment processing on the mono recording data before the actual communication analysis begins. By pre-separating the customer and agent voice streams and synchronizing them temporally, the system enables subsequent analysis operations to proceed quickly on already-separated data, achieving both speed and accuracy in the overall analysis process.
Data Source
AI summary
The methods, apparatus, non-transitory computer readable media, and systems described herein include recording a mono recording of a software and a customer inquiry communication using a microphone to interpret and respond to the customer inquiry communication, wherein the mono recording is unseparated and includes customer voice data and audio data generated by the software agent, separately and concurrently recording the software agent audio data in an agent recording, and subtracting agent audio data from the unseparated mono recording to provide a separated recording including only customer voice data.