Can your voice reveal whether you have an illness? 闻声探病是真是假?

2017-01-19 11:46 千龙网

打印 放大 缩小

Our voices are amazing things.

我们的声音很奇妙。

We can use them to sing, shout and whisper sweet nothings. We can use them to activate gadgets and prove who we are to banks.

它可以是一段歌唱、一阵喊叫、或是绵绵细语,它也可以是激活设备的一道口令,或是我们在银行的身份证明。

And now researchers believe they can also reveal whether we're getting ill.

而目前研究人员发现声音还能够透露我们是否患有疾病。

A US start-up called Canary Speech is developing a way of analysing conversations using machine learning to test for a number of neurological and cognitive diseases, ranging from Parkinson's to dementia痴呆.

一家名叫Canary Speech的美国初创公司正在研究通过分析对话筛查疾病的方法。他们通过机器学习技术对受测人的对话加以分析,以此检查出包括帕金森病和痴呆在内的神经和认知方面的疾病。

The project was born out of a painful personal experience for the firm's co-founder Henry O'Connell.

该项目源自公司联合创始人亨利•奥康奈尔的一段苦痛往事。

"It has been my pleasure to have as a friend for nearly 30 years a dear gentleman who was diagnosed six years ago with Parkinson's disease," says Mr O'Connell.

“我和一位朋友有近三十年的交情,他是一位和蔼可亲的绅士,能和他成为朋友是我的荣幸。六年前,他被诊断出帕金森病”, 奥康奈尔说。

"My friend was told when the diagnosis was finally made that it was likely that he had been suffering from Parkinson's for over 10 years."

“在确诊时,医生告诉我的朋友他可能在十多年前就已经患有帕金森病了。”

As with so many diseases, early diagnosis can play a crucial role in effectively managing the condition, but recent research highlights the difficulties in correctly diagnosing it, with doctors often struggling to distinguish the symptoms.

与很多疾病的防治一样,及早发现病情对于有效地控制帕金森病非常关键,但最近的研究显示正确诊断仍然存在诸多困难,医生经常难以通过症状来确诊。

And the longer the condition goes undiagnosed, the more severe the symptoms become.

而疾病确诊越晚,症状就会变得越严重。

"During the years before his diagnosis was accurately made, my friend, suffering from muscle and apparent nerve-related pain, was treated in several medical facilities," says Mr O'Connell.

“在我的朋友被确诊之前的那些年,他一直遭受着肌肉和神经方面的疼痛,并去了好几家医院”,奥康奈尔说道。

"The muscle and nerve-related pain were directly associated with a progressing Parkinson's illness. Because it went undiagnosed, proper treatment was delayed and his Parkinson's progressed potentially more rapidly than it would have under proper diagnosis and treatment."

“肌肉和神经疼和帕金森病情的加重直接相关。由于没有确诊,他的帕金森病一直没有得到正确的治疗,要是他的病情能够早点发现、对症治疗的话,就不会发展得那么快。”

Canary Speech developed algorithms after examining the speech patterns of patients with particular conditions, including Alzheimer's, dementia and Parkinson's.

在研究了一些包括患有阿尔茨海默症、痴呆和帕金森病的特殊病人的语言模式后,Canary Speech开发了一种软件。

This enabled them to spot a number of tell-tale signs 迹象 both pre and post-diagnosis, including the kinds of words used, their phrasing语调, and the overall quality of speech.

这种软件可以发现这些疾病在诊断前后的许多迹象,其中包括在病人的措辞、语调和整体表达能力方面的迹象。

For instance, one symptom of the disease is a softening of the voice - something that can be easily overlooked by those close to us. But Canary Speech's software is capable of picking up such small changes in speech patterns.

例如,疾病的一大症状是声音变轻,这点很容易被忽视,而Canary Speech的软件能够捕捉到言谈中的这种细微变化。

The overall goal is to be able to spot the onset of these conditions considerably sooner than is currently possible. In initial trials, the software was used to provide real-time analysis of conversations between patients and their clinicians.

Canary Speech研发软件的总体目标是大大提前发现疾病的初期症状的时间。在最初的测试中,软件被用来为病人和临床医生之间的谈话提供实时分析。

As with so many machine learning-based technologies, it will improve as it gains access to more data to train the algorithms that underpin 支持 it.

和许多基于机器学习的技术一样,软件随着获得的数据量的增加会得到不断进化。

And as more voice-activated devices come on to the market and digital conversations are recorded, the opportunities to analyse all this data will also increase.

并且随着越来越多的声感设备进入市场和越来越多声音被数字化,软件可获得的分析数据也会增加。

Some researchers have analysed conversations between patients and drug and alcohol counsellors, for example, to assess the degree of empathy the therapists were displaying.

所以一些研究者才得以通过分析病人与药物和酒精顾问之间的对话来评估顾问在情感上与病人的共鸣程度。这是软件进化的一个例证。

"Machine learning and artificial intelligence has a major role to play in healthcare," says Tony Young, national clinical lead for innovation at NHS England.

英格兰国民卫生服务机构的国家临床创新部负责人托尼•杨说:“机器学习和人工智能在医疗保健中起着至关重要的作用。”

"You only have to look at the rapid advancements made in the last two years in the translation space. Machine learning won't replace clinicians, but it will help them do things that no humans could previously do."

“你只要看看最近两年机器学习在翻译领域里取得的迅猛进步就会明白,机器学习技术不但不会取代临床医生,而且能帮助他们去完成以前没人能完成的事情。”

It is easy to see how such technology could be applied to teaching and training scenarios.

不难看出这种技术同样可以应用于教学和培训中。

How's my talking?

我聊得怎么样?

Voice analysis is also being used in commercial settings.

目前,声音分析技术也被应用于商业活动。

For instance, tech start-up Cogito, which emerged from Massachusetts Institute of Technology, analyses the conversations taking place between customer service staff and customers.

麻省理工学院科技初创公司Cogito对话务员和来电者之间的对话分析即是一例。

They monitor interactions in real time. Their machine learning software compares the conversation with its database of successful calls from the past.

他们对双方的互动进行了实时监测,并利用机器学习软件与数据库中的成功电话对话进行了对比。

The team believes that it can provide staff with real-time feedback on how the conversation is going, together with advice on how to guide things in a better direction - what it calls "emotional intelligence".

他们认为声音分析技术可以为话务员提供谈话进展情况的实时反馈,同时还可以为其提供建议,引导谈话向更好的方面发展——即所谓的培养“情商”。

These tips can include altering one's tone or cadence to mirror that of the customer, or gauging the emotions on display to try to calm the conversation down.

该技术会根据来电者的语气和说话节奏向话务员提出调整说话语气和节奏等的建议,还可以监测话务员流露出来的情绪,并提醒其尽量保持平和地交谈。

It's even capable of alerting the supervisor if it thinks that greater authority would help the conversation reach a more positive conclusion. The advice uses the same kind of behavioural economics used so famously by the UK government's Behavioural Insights Team, also known as the Nudge Unit.

声音分析技术甚至还具备判断能力,在它认为某一通电话需要话务主管介入才能取得更积极的结果时,它就会提示话务主管。它给出建议的依据与英国政府的行为研究小组(也叫“推一把小组”)的依据相同,即大名鼎鼎的行为经济学。

Early customers of Cogito's product, including Humana, Zurich and CareFirst BlueCross, report an increase in customer satisfaction of around 20%.

通过使用Cogito的产品,瑚玛娜、苏黎世和CareFirst BlueCross等一批早期用户的客户满意度都增张了约20%。

As the internet of things spreads its tentacles触须throughout our lives, voice analysis will undoubtedly be added to other biometric ways of authenticating ourselves in a growing number of situations.

随着物联网在人们生活中应用得越来越广,声音分析技术也必将被广泛应用到越来越多采用计量生物学方法来进行身份验证的领域。

Google's Project Abacus, for example, is dedicated to killing passwords, given that 70% of us apparently forget them every month.

谷歌的“算盘项目”就是一个例子。70%的人每个月都会忘记密码,而“算盘项目”就是为了免去密码输入而生的。

It plans to use our speech patterns - not just what we say but how we say it - in conjunction with other behavioural data, such as how we type, to build up a more reliable picture of our identity. Our smartphones will know who we are just by the way we use them.

该项目计划利用我们的语言模式——这不光包括我们说话的内容,还包括我们说话的方式——以及打字习惯等的行为数据,建立一个更可靠的身份验证体系。这样,一部智能手机就可以通过使用者的使用方式识别他的身份。

The big - silent - elephant in the room is how all this monitoring and analysis of our voices will impact upon our right to privacy.

但是这些对声音的监测和分析会对隐私权产生何种影响呢?这个问题显而易见,但却无人问津。

注:英文资料来源http://www.bbc.com/news/business-38637257

责任编辑:Lu Qing(QN0045)