Shulou(Shulou.com)06/02 Report--

The first time I used a mobile voice assistant, this strange thing always answered disappointingly after I excitedly asked a question, "I really want to understand what you're talking about."... "

Later, with the evolution of technology, it was finally able to understand what I said through some of my keywords. But one by one, word by word, stylized.

Slowly, it can no longer be overcome by a simple conversation. At the beginning, the speech assistant is not limited to the understanding of the "text", but also adds visual, auditory and other analysis techniques to capture the emotion behind people's "discourse" through "observing words and colors". They began to try to capture human emotions.

As Amazon recently revealed, it has plans to build a watch-like wearable device to provide more services by capturing users' emotions, based on the technology's ability to recognize emotions in Alexa.

Everything is inseparable from "audio-visual": the main way for AI to detect emotions

The research and summary of human emotions is a topic that has been carried out for nearly a century and a half. At first, researchers traveled to and from different ethnic groups, compared and analyzed, and obtained several basic emotional categories of human beings, which can then be used in interpersonal communication, medical care and other industries.

Today, with the deepening of human-computer interaction, whether the machine can fully understand human emotions will directly affect the interactive experience. As a result, many artificial intelligence companies began to tackle the problem of machine recognition emotion. Thanks to the continuous breakthroughs in image recognition and speech recognition technology, there are two main ways for AI to recognize emotions.

The first is visual emotion recognition. This is the most basic way of emotion recognition, because human emotions are often expressed directly in the form of facial muscle movements, which is the direct reason why American psychologists Paul Ekman and Follison have proved that human beings have six basic emotions.

Facial recognition is now a very mature technology, by "feeding" AI a large amount of facial expression data, and then learning, so that we can judge the emotions of real faces. In China, there are artificial intelligence companies that specialize in facial emotion recognition, such as amblyopia, Swire and so on.

The second is speech emotion recognition. People's emotions can be "seen" and "heard". What is directly related to it is the tone of speech in different emotions: deep in sadness, high in pleasure, and so on.

Compared with visual recognition, speech recognition is much more difficult. Because the correspondence between emotion and expression is much higher than that of intonation and emotion, for example, it is difficult to associate a sad face with happiness, but if a person speaks in a low voice, it can indicate a heavy heart, or it may be because of his timbre. Because of this, the analysis of sound signals is required to be more precise. Baidu, Google, Amazon and so on all have a certain amount of technology accumulation.

In addition to these two ways of emotion recognition, text recognition and physiological signals (brain waves, respiration, skin resistance, etc.) are also methods of emotion recognition by AI, but because of the narrow scene and low accuracy of physiological signals, it can not become the mainstream emotion recognition method.

Obviously, what Alexa is good at is the second approach. In terms of speech emotion recognition, Alexa patented it last October to determine users' happiness, happiness, anger, sadness, fear, disgust, boredom, stress and other emotions through voice patterns. Now, the patent is finally coming in handy.

From buying cough syrup to online shopping: what happens when Alexa understands emotions

Alexa is Amazon's team that specializes in intelligent voice technology, and its technology transformation is often related to another Amazon team, Lab 126, which has developed products such as kindle and echo. This time, the two teams are once again working together to put the latest results of voice emotion recognition on a wearable device similar to a smartwatch. Through the connection with the mobile phone APP, to provide personalized services for users.

According to Amazon's internal documents, the function of the watch is simple: health assistance. In addition, it can also identify the voice characteristics of users to make emotional judgment, and then provide them with the best way to communicate with others.

According to the description, if a girl blows her nose and coughs and tells Alexa that she is a little hungry, then Alexa will analyze the tone (weak, deep) and background sound (cough, blowing nose) of talking to the girl that she is probably sick, and then send out a thoughtful care from the machine: would you like a bowl of chicken soup or order a takeout? Even order a bottle of cough syrup directly online and deliver it to your door within an hour?

To tell you the truth, if a girl really has such a sweet assistant, what kind of boyfriend does she need?

From this perspective, Alexa's role as a personal life assistant will be further strengthened. For example, connect it with mobile music, e-commerce APP and other businesses, sing you a happy song when you feel sad, or buy you a box of desserts to make you feel happy through food.

As a result, Alexa is likely to open up a new way of voice shopping recommendation. Unlike the traditional method of capturing users' browsing records, search records or "stealing" the privacy footprints left by other users on other APP, and then making aimless generalization and inefficient recommendations, shopping recommendations for users through perceptual emotions and inquiries can achieve higher transaction efficiency. This will undoubtedly form a great attraction to e-commerce, and it is also likely to become a new business model for Alexa to make a profit.

Of course, this wearable device is still in the early stages of research. Amazon is not entirely sure whether to implement the plan. If the technology is feasible, the design will be carried out further. We seem to be "over-reading" in such a long speech.

But in this day and age, the most important and unpredictable thing is the trend of technology. If it does come into reality, there may be something that needs to be taken seriously and dealt with properly.

Beware of turning "service" into "business"

The first thing that needs to be paid attention to is the privacy issue that we take the trouble to mention again and again but have to mention.

Private data is unavoidable in the Internet era. Users provide their own data to obtain services. This exchange takes place every day. In all kinds of APP, in order to improve the user experience of the product, the service provider generally explicitly tells that it will collect part of the user's privacy data, and it is only used to improve the product. But this has always been regarded as nonsense, especially since Alexa still has a criminal record-echo has been exposed more than once by users for eavesdropping and privacy collection.

But the situation is different if it is made into a wearable device to monitor health and identify emotions. Under normal circumstances, smart speakers need to be awakened to work (if abnormal conditions do not exist …...) On the other hand, wearable devices such as smartwatches need to be turned on 24 / 7, which means they are aware of your changes all the time. Come to think of it, there is something that silently controls everything about your emotions, and then it is likely to do something unknowingly behind your back, isn't it a little harmful to pia?

Where is the boundary of obtaining privacy? it is the first problem that needs to be solved.

Secondly, we need to guard against its "alienation" from a private service to an unreasonable commercial use. As we have analyzed above, it can reach a cooperation with e-commerce to form a new shopping model. But what if the goods or merchants it recommends are "recharged"? You think the recommended product is the most suitable, but unexpectedly it is the most appropriate to give money. Taking advantage of users' emotions for commercial distribution will obviously add a lot of congestion to users.

From this point of view, it is also necessary to focus on how to draw the line between service and business.

Finally, every time you notice a change in a person's mood, you will respond in a timely manner, which is likely to have two diametrically opposite effects on the person.

One is to rely too much. Bad mood has its appeasement, emotional pleasure has its "icing on the cake", will let people follow nature, lack of independent emotional regulation ability. Adults are fine, if they are children, it is likely to have a negative impact on the formation of their character. How to restrict children's use of such products will become a new problem.

The second is to arouse revulsion. People have different personalities, some people may like to be pacified, but some people will be irritated if they are "booed and asked for warmth" in every detail. This is easy to understand, just like your parents' nagging is the same reason. So, how to master the rhythm that can provide intimate service by detecting emotions without causing disgust is also a technical job.

Generally speaking, Alexa is worthy of recognition if it can provide positive services based on emotion recognition. After all, it is inevitable to have emotional collisions in the final stage of human-computer interaction, and it is understandable to provide some commercial services based on this. However, due to the privacy and personalization of emotions, both Alexa and other companies that provide similar voice technology services must provide measured and borderline services on the premise of ethical common sense and guaranteed experience.

Otherwise, it will be doomed to be abandoned by the user.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.