Same as Fingerprint, Voiceprint is a unique biological feature of people, could be used to recognize “Who are you”.

Comparing with face recognition and fingerprint recognition, voiceprint recognition has obvious advantages such as low sensitive, low cost, and easier to capture. Herewith, it has been widely used on non-contact or remote speaker identification scenarios.

Based on self-developed voiceprint recognition technologies, Junlin is providing more convenient and safer ID recognition solutions to customers in different fields.

Voiceprint Recognition Voice Wake-up Voice Expansion Voiceprint Database

Based on self-developed deep learning algorithm and integrated with self-developed acoustic technologies, the accuracy and stability of Junlin voiceprint recognition are in the international top level already. The recognition types include Random Figures, Text Dependent, Text Independent and No-registration voiceprint.

Recognition types and the Accuracy Rate (natural environment)


Random Figures

Voiceprint recognition with Random Figures requests users to register their voiceprint by repeating the random figures from 0 to 9 according to the system requirement, and then complete the voiceprint recognition by speaking the random figures from 0 to 9 provided by the system.


Text Dependent

Voiceprint recognition with Text Dependent requests users to register the voiceprint by repeating the fixed text defined by the system, and then complete the voiceprint recognition by speaking the registered fixed text.


Text Independent

Voiceprint recognition with Text Independent means users can decide their own texts for the voiceprint registration, and then register the voiceprint by repeating the text they defined, and then complete the voiceprint recognition by speaking the registered customized text.


No-registration Voiceprint

One of the few companies who are developing this technology. A voiceprint recognition system doesn’t need registration. The system can collect the voice of different users, recognize their voiceprints, and complete the registration automatically.

Technical Advantages

High accuracy equal error rate is below 1%

Self-developed Signal processing and acoustic feature extraction algorithm. Under laboratory environment, the FAR of the 1:100,000 recognition is below 1%.

Millisecond level response time

Complete the voiceprint recognition instantly, no lag feeling for users.

High robustness, stable recognition rate

Strong stability of time, the accuracy of the model won’t be influenced by time lapses.

Based on cloud platform and vessel technology, easy for extending

Could adapt different systems such as Linux, Windows, Android, and meet the expanding needs of the explosive data increase.

Two models includes SDK and API

Able to be integrated locally, or be deployed on private/public cloud according to user needs. Also easy for the cross-platform integration.

Technology Demonstration