文 | By Chai Hua . CHINADAILY HongKong
原文 | Voice recognition likely to be an alternative to passwords
譯 | Minya . VoiceAI Technologies
Editor’s note:The need for better cybersecurity is still usually addressed with passwords, a flawed, antiquated technology that most computer users find frustrating. The need for an easier yet secure alternative has one company focusing on the human voice as a foolproof alternative.編者註：對於網絡安全的更高需求，目前仍通過使用密碼這種令大多數用戶沮喪、缺陷滿滿的過時技術來解決。由於對更簡單且更安全的替代方案的需求，令一家公司專注於研究使用人類的聲音創造一個萬無一失的替代方案。
Passwords are quietly falling out of vogue, as frustration with setting up and remembering complicated codes has become a barrier to attracting and retaining users of apps and websites.
用戶在設置和記憶複雜密碼時的煩惱，已成為應用和網站吸引和留存用戶的障礙，這使得密碼正在悄然被淘汰。
About 85 percent of consumers forget at least one password a year, according to Gygya, a customer identity management platform. In addition, a fourth of people interviewed said they don’t create complicated passwords for their financial accounts. More than half of the consumers said that if they were given the choice, they would do away with user names and passwords altogether in favor of other forms of authentication。
根據客戶身份管理平臺Gygya的說法，大約有85％的消費者每年至少忘記一個密碼。 此外，四分之一的受訪者表示，他們不會為自己的金融帳戶創建複雜的密碼。 超過一半的消費者表示，如果給予他們選擇，他們支持其他形式的身份驗證來完全取代傳統的用戶名和密碼。
Amid this dissatisfaction with the status quo, biometric authentication — using a person’s unique physical characteristics to verify one’s identity — is emerging as a preferred alternative.
在對現狀不滿的情況下，生物識別身份驗證——即使用一個人獨特的生理特徵來驗證一個人的身份——正在成為首選替代方案。
Our company’s AI-empowered system can compare a voice to 1 million others in one second.我們的AI系統可以在1秒鐘內將1條聲音的聲紋與100萬條聲紋進行比對——Li Yatongfounder of VoiceAI Technologies Co.聲揚科技創始人李亞桐」
One of the most sought-after biometric authentication methods now is facial recognition, but a startup in Shenzhen is taking another route to integrate artificial intelligence and voiceprint technology.
現在最受歡迎的生物識別認證方法之一是人臉識別，但深圳的一家初創公司正將人工智慧和聲紋技術結合在一起作為另一種可行的方式。
Among biometric authentication methods, including fingerprint scanning, facial recognition and iris scanning, voice recognition ranked at the second-highest preference, Gygya’s survey found.
Gygya的調查發現，在包括指紋識別、人臉識別和虹膜識別等各種生物識別認證方法中，聲紋識別是第二受歡迎的。
According to another survey by Opus Research in 2016, more than 135 million people worldwide had enrolled for services that would use their respective voices to speed up the authentication process.
根據Opus Research在2016年進行的另一項調查，全球已有超過1.35億人註冊並使用自身聲音特徵來加速身份驗證過程。
Development of the technology has also received official support in China. The central bank issued technical specifications in October for voiceprint-recognition-based security applications related to mobile finance, setting up industry standards in the sector.
該技術的發展也得到了中國的官方支持。 2018年10月，中國人民銀行發布了《移動金融基於聲紋識別的安全應用技術規範》，建立了金融領域應用聲紋識別技術的行業標準。
In fact, voiceprint recognition is not a new technology, and many countries have applied it to criminal investigations, said Li Yatong, founder of VoiceAI Technologies Co. He said investigators relied on human’s hearing and sound spectrographs to judge identity in the past.
事實上，聲紋識別並不是一項新技術，許多國家已將其應用於刑事調查，聲揚科技的始人李亞桐表示，「過去，調查人員依靠人類的聽力和聲譜圖來鑑定身份。」
Now as AI technology develops rapidly, traditional voiceprint recognition methods have also remarkably improved. 「Our AI-empowered system can compare a voice to 1 million others in one second,」 Li said.
「現在隨著人工智慧技術的迅速發展，傳統的聲紋識別方法也得到了顯著的改善。」李亞桐說，「我們的AI系統可以在1秒鐘內將1條聲音的聲紋與100萬條聲紋進行比對。」
More importantly, he also promises it can distinguish a live human voice from recorded sound, based on its unique algorithm. He also pledges the accuracy rate of the technology can reach 99.5 percent, while the entry-level accuracy for biological recognition commercial application is 95 percent.
更重要的是，他還表示，基於其獨有的算法，現場人聲與錄製聲音可以區分出來。 他還提到該技術在實際應用中的準確率已達到99.5％，而市面上公認的生物識別技術可商用的準確率需要至少達到95%。
Li said he believes the technology has a wide range of applications, including in mobile phone software, robots and internet-connected cars.
李亞桐說他相信這項技術有著廣泛的應用場景，包括手機軟體、機器人和車聯網。
Another trend is that multiple methods of biometric authentication will be used simultaneously to guarantee safety, he added.
未來的另一個趨勢是將同時應用多種生物認證方法來更好地保障安全性，李亞桐補充道。
Last year, VoiceAI provided a customized system for the Indonesian government to recognize legal claims to pensions, in a project covering 2.5 million retired civil servants in the nation.
去年，聲揚科技為印度尼西亞政府定製了養老金認證系統，該項目覆蓋印尼250萬名退休國家公務員。
Previously, elderly residents were required to go to banks or retirement institutions to collect their pensions in order to prevent false claims, but now they can choose any two of fingerprint photo, face recognition or voiceprint recognition to make claims from home.
在此之前，為了防止虛假申領，老年人必須到銀行或退休機構現場領取養老金，但現在他們可以選擇指紋照片、人臉識別或聲紋識別中的任意兩種方式在家中遠程認證來申領養老金。
For voiceprint recognition, which Li said is the most favored option, the users need only read aloud several numbers.
李亞桐說，在這個項目上線後，聲紋識別是其中最受歡迎的選項，用戶只需要大聲讀出幾個數字就可以完成聲紋認證。
A Mexico national finance institution also adopted VoiceAI’s technology in the same year. The startup now provides identity recognition services for the institution’s social security, mortgage and loan products. It also teamed up with a company to develop vehicle-mounted systems to integrate voiceprint recognition into private cars. In the future, drivers will easily be able to use their voiceprints to start their cars, log onto applications, or make payments.
同年，墨西哥一家國家級金融機構也採用了聲揚科技的技術。這家初創公司現在為該機構的社保、按揭和貸款產品提供基於聲紋的身份識別服務。該公司還與一家公司合作開發車載系統，將聲紋識別技術集成到私家車中。在未來，司機可以很容易地使用他們的聲紋啟動他們的汽車、登錄應用程式或進行支付。
Li also disclosed several domestic financial institutions are in negotiation to apply VoiceAI’s technology.
李亞桐還透露，若干家國內金融機構正在與聲揚科技洽談採用其技術。
One possible application is to double-guarantee information security with a verification code, which now is widely regarded as a simple supplementary log-on method to personal passwords, especially internet-based wealth management platforms.
使用驗證碼是一種雙重保證信息安全的可行方式，目前，簡訊驗證碼被廣泛認為是個人密碼的簡單補充登錄方式，尤其是基於網際網路的財富管理平臺。
But it also makes stolen phones vulnerable. VoiceAI’s voiceprint solution is to identify users』 authenticity by asking them to read the verification code aloud.
但簡訊驗證碼也令手機被偷帶來的風險變得更大。針對這個問題，聲揚科技通過讓用戶朗讀驗證碼（從而同時驗證其聲紋和朗讀內容的正確性）的方式來識別用戶的真實身份。
In the last two years, Li’s startup has achieved an annual growth of 10 times in revenue, according to Li.
據李亞桐說，在過去的兩年裡，公司的年收入增長了10倍。
He attributed the fast growth to his elite team. In 2016, a trio of top AI-based identity recognition experts founded the startup and received an investment of 10-million level yuan at the latest round of funding in February. Li himself has a Master of Business Administration degree from The Chinese University of Hong Kong and has 10 years of working experience in the identity safety and biological recognition industry. Another two founders, Zhang Weibin and Chen Dongpeng, are both doctors in the field of AI and voice at Hong Kong University of Science and Technology.
他將這一快速增長歸功於他的精英團隊。 2016年，三位頂級人工智慧身份識別專家創辦了聲揚科技，並在2019年2月份的最新一輪融資中獲得了千萬級人民幣的投資。 李亞桐自身除了擁有香港中文大學工商管理碩士學位，還在身份安全和生物識別行業擁有10多年的工作經驗。 另外兩位創始人張偉彬和陳東鵬都是香港科技大學人工智慧和語音領域的博士。
With 70 percent of his 30-some workmates in research and development, Li pays extra attention to technology development. He is now discussing with Hong Kong universities the possibility of setting up a joint laboratory for fundamental research and application development of voiceprint recognition and 「intelligent speech,」 which includes speech recognition, semantic analysis and voice synthesis — the entire spectrum of human-to-machine communication.
李亞桐對技術投入了非常多的精力和資源，他的30多名同事中有70％從事研發。 目前他正與香港的高校討論合作建立聯合實驗室的可能性，來進行聲紋識別和「智能語音」——包括語音識別、語義分析和語音合成等全棧人機通訊技術——方面的基礎研究和應用開發。
Similar to the development path of face recognition technology, talented employees are the most valuable asset of the industry now, Li said.
李亞桐說，與人臉識別技術的發展路徑類似，人才是目前行業中最寶貴的資產。