Baidu announced the opening of 4 voice interfaces: claims to revolutionize human-computer interaction-IT information
On November 22 the message, Baidu announced today the opening of four new speech technology interfaces, help users enhance the interactive experience when you use speech to communicate with machines, which coincided with the open voice platform of Baidu launched 3 years ago.
Baidu's Chief Scientist, Wu Enda
Baidu's Chief Scientist, Wu Enda introduced the four voice technology, emotion synthesis, respectively, far-field programme, the arousal phase II technology and long audio programme.
"These technologies have great potential to revolutionize the efficiency and human-computer interaction approach. Voice technology of the future in many application scenarios have a very good chance, will bring big changes in human-computer interaction. "Wu Enda said.
Specific aspects of these technologies were designed to solve user interaction using speech when General bothers a number of key issues. For example, Baidu emotion focused on speech synthesis technology "added emotion", currently available to nearly live voice effects, Baidu earlier this year using this technology, rehabilitation has gone star Leslie Cheung's voice.
Similarly, developers can also take advantage of the new interface, speech recognition range increased to 3-5 metres, to wake device voice rate up to 95% per cent and far more power-less false positives, or enhance the accuracy of speech recognition problem for a long time. This voice technology will bring far more imagination than it is now, and not just a remote control TV, or unlock the phone.
For example, the first two are representatives of Baidu's "little robot man-machine speech interaction order" flagship store in Shanghai KFC put into application, remote answering at any time to order. The latter has many records, intelligent service, video transfer application context.
Follows the resignation speech open platform has been in mobile phones, home appliances, automotive and communications services, and other areas of cooperation covered, specific partners, including Lenovo, ZTE, CHANGHONG, KONKA, SONY, Tesla, Tucson, Hewlett-Packard, Amy communications, ctrip, mobile QQ and reading.
Public information, Baidu open voice platform launched in October 2013, the current daily online speech recognition requires 140 million, online speech synthesis daily requests reached 200 million, developer number more than 140,000.
It is worth mentioning that, under Wu Enda and its research and development team, Baidu speech recognition accuracy rate has been as high as 97%, deep Deep Speech voice recognition system 2 also named MIT 20,160 breakthrough technologies.
百度宣布开放4项语音技术接口：号称彻底改变人机交互方式 - IT资讯
11月22日下午消息，百度今日宣布开放四项全新语音技术的接口，帮助用户提升在使用语音与机器进行交流时的交互体验，该决定恰逢百度语音开放平台上线3周年。
百度首席科学家吴恩达
百度首席科学家吴恩达介绍了这四项语音技术，分别是情感合成、远场方案、唤醒二期技术和长语音方案。
“这些技术有很大的潜力，去彻底改变人机交互的效率和办法。未来语音技术在很多应用场景有很好的机会，将为人机交互带来巨大的改变。”吴恩达表示。
具体方面，这些技术旨在解决用户在使用语音交互的场合时，普遍感到困扰的一些关键问题。例如，百度情感合成技术主要聚焦在为合成语音“加入情感”，目前可达到接近真人发声效果，百度今年早些时候曾利用此技术，复原已逝明星张国荣的声音。
类似地，开发者还可以利用新的接口，使语音识别距离增加到3-5米，将设备的语音唤醒率提升到95%以上同时更省电误报更少，或提升长时间语音识别的准确率问题。这将为语音技术带来远比现在更多的想象空间，而不只是遥控电视或解锁手机。
例如，前两者的代表是百度的“小度机器人人机语音交互点餐”，已在上海肯德基旗舰店投入应用，可远距离随时应答点餐。后者则已经在诸多内容记录、智能客服、视频转写等场景实现应用。
此前，百度语音开放平台已经在手机、家电、汽车和通讯服务等多领域实现合作覆盖，具体合作伙伴包括联想、中兴、长虹、康佳、SONY、特斯拉、途胜、惠普、艾米通讯、携程，及手机QQ阅读等。
公开资料显示，百度语音开放平台上线于2013年10月，目前每日在线语音识别要求1亿4千万，在线语音合成每日请求达2亿，开发者数量超过14万。
值得一提的是，在吴恩达及其团队的研发下，百度语音识别准确率目前已高达97%，深度语音识别系统Deep Speech 2还入选了MIT 2016十大突破技术。