Real-time Voiceprint Identification in Call Centers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing voiceprint identifying technologies in call centers face hysteresis issues due to the requirement for sound duration and the inclusion of operator sounds in recordings, leading to increased identification time.
Innovation Solution
Sending voice information directly to a voiceprint identifying device in real-time through a voice site channel without recording, using a pre-set duration of 8 seconds, to reduce the time required for user identity identification and minimize hysteresis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sound recording is used for voiceprint identification, then user identity can be identified, but identification time increases due to hysteresis and operator sound inclusion
Solution Approach 1:
The patent extracts only the client's voice information from the mixed audio stream, separating it from the operator's voice. This is achieved by sending only the extracted client voice data to the voiceprint identification device, eliminating the need to process entire recordings containing both parties' voices.
Solution Approach 2:
The patent performs preliminary voice information extraction and prepares the client's voice data before it is mixed with the operator's voice in the recording. By extracting and processing the client's voice information in advance, the system avoids the time delay that would occur if extraction had to happen after recording.
2Measurement precision
If sound recording is required for voiceprint identification, then identification can be performed, but the process includes unnecessary operator sounds
Solution Approach 1:
The patent extracts only the necessary client voice information from the audio stream, removing the operator's voice and other unnecessary components. This extraction process sends only the relevant voice data to the voiceprint identification device, simplifying the processing while maintaining identification accuracy.
3Speed
If voice information is sent through voice site channel, then real-time transmission is achieved, but infrastructure complexity increases
Solution Approach 1:
The patent utilizes the existing voice site channel infrastructure, which is already part of the call center system, to transmit voice information. This approach leverages the multi-functionality of the voice site channel, which can handle both voice communication and voiceprint data transmission, thereby avoiding the need for separate dedicated transmission infrastructure.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A user identity identifying method is provided, which includes the following steps: voice information of a client joining in a voice site of a call center is sent to a voiceprint identifying device by using a channel of the voice site; a voiceprint identification result returned by the voiceprint identifying device is received, and the user identity corresponding to the voice information of the client is validated according to the voiceprint identification result. A user identity identifying device and call center system are further provided, in which voice signals of a client are sent to a voiceprint identifying device in real time through a channel of a voice site without the need of recording sound signals of the client, and therefore, the time required for the user identity identification is reduced, and the hysteresis in the user identity identification is decreased effectively.