In the process of current biometric technology development, information such as faces and fingerprints may be easily stolen. As an important biological feature that is not easy to copy, voiceprint makes the advantage of voiceprint recognition technology very obvious in an unsupervised environment. Speech is a simple signal with rich contents. Thus the voiceprint recognition is the best solution in authentication. Voiceprint recognition is an evolving blue ocean market emerging so many interesting participants.
Voice AI Technologies Co., Ltd completed ten million-Level pre-A financing in 2019. D-Ear Technologies led by the professor Zheng Fang from Tsing Hua has won large customers such as China Contribution Bank, National Information Center, and China UnionPay. The most popular smart speakers in recent years have also added the voiceprint recognition part. Voiceprint Recognition means that it can identify the speaker only by sound. Everyone will differ in language habits, sound volume and frequency. Like fingerprint, each voiceprint is unique thus can be used to identify the personal ID.
With its rapid development, voiceprint recognition technology has gradually applied in financial industry and social software. As early as the first half in 2015, the Wechat APP has released the ‘voice lock’ and appealed many users to have a try. It can easily verify your identity only based on the voice waveform, and nothing to do with accent or language.
At present, voiceprint recognition technology has
achieved great results in certain topics (like digital strings) and 1:1
verification, but it still faces many technical challenges in open topics and
1:N searching. Plenty of enterprises, institutions are still working hard in
this field to search a broader application scenario.
The development of technology depends on large quantity of high-quality data. Following the trend of AI research and development, Datatang provide all kinds of voiceprint data to assist the research and development of voiceprint recognition technology.
In terms of certain topics, 10,000 speakers record Chinese digital strings speech data. Each speaker records 30 digital strings, covering 4 digitals, 6 digitals and 8 digitals.
In terms of open topics, Datatang collected from 500 speakers, with each speakers 300 speech & texts data. In order to reduce the variability of the collected person's voice and increase the diversity of the same person's voice, each person is recorded four times at intervals of two to three weeks.The change of technology has brought about changes in lifestyle. When all of us have our own unique voiceprint "password", the traditional Pin passwords and fingerprints may be replaced. The rapid development of technology requires a large amount of high-quality data, and DATATANG is willing to be the solid and reliable data anchor behind the research and development of artificial intelligence products.