Crossing the Uncanny Valley: Breakthrough in technology for lifelike facial expressions in androids

穿越恐怖谷：机器人中逼真面部表情的技术突破

A research group led by Osaka University has developed a technology that allows androids to dynamically express their mood states, such as "excited" or "sleepy," by synthesizing facial movements as superimposed decaying waves.

Even if an android's appearance is so realistic that it could be mistaken for a human in a photograph, watching it move in person can feel a bit unsettling. It can smile, frown, or display other various, familiar expressions, but finding a consistent emotional state behind those expressions can be difficult, leaving you unsure of what it is truly feeling and creating a sense of unease.

Until now, when allowing robots that can move many parts of their face, like androids, to display facial expressions for extended periods, a 'patchwork method' has been used. This method involves preparing multiple pre-arranged action scenarios to ensure that unnatural facial movements are excluded while switching between these scenarios as needed.

However, this poses practical challenges, such as preparing complex action scenarios beforehand, minimizing noticeable unnatural movements during transitions, and fine-tuning movements to subtly control the expressions conveyed.

In this study, lead author Hisashi Ishihara and his research group developed a dynamic facial expression synthesis technology using "waveform movements," which represents various gestures that constitute facial movements, such as "breathing," "blinking," and "yawning," as individual waves. These waves are propagated to the related facial areas and are overlaid to generate complex facial movements in real time. This method eliminates the need for the preparation of complex and diverse action data while also avoiding noticeable movement transitions.

Furthermore, by introducing "waveform modulation," which adjusts the individual waveforms based on the robot's internal state, changes in internal conditions, such as mood, can be instantly reflected as variations in facial movements.

"Advancing this research in dynamic facial expression synthesis will enable robots capable of complex facial movements to exhibit more lively expressions and convey mood changes that respond to their surrounding circumstances, including interactions with humans," says senior author Koichi Osuka. "This could greatly enrich emotional communication between humans and robots."

Ishihara adds, "Rather than creating superficial movements, further development of a system in which internal emotions are reflected in every detail of an android's actions could lead to the creation of androids perceived as having a heart."

By realizing the function to adaptively adjust and express emotions, this technology is expected to significantly enhance the value of communication robots, allowing them to exchange information with humans in a more natural, humanlike manner.

dynamically [daɪˈnæmɪkli] adv. 动态地；有活力地

synthesize [ˈsɪnθəsaɪz] v. 合成；综合

superimpose [ˌsuːpərɪmˈpəʊz] v. 叠加；放在……上面

unsettling [ʌnˈsetlɪŋ] adj. 令人不安的；扰乱的

consistent [kənˈsɪstənt] adj. 始终如一的；一致的；相符的

scenario [səˈnɑːriəʊ] n. 设想；方案；情节；剧本提纲

pose [pəʊz] v. 造成（威胁、问题等）；提出（问题等）；摆姿势；n. 姿势

subtly [ˈsʌtli] adv. 微妙地；巧妙地；细微地

constitute [ˈkɒnstɪtjuːt] v. 组成；构成；被视为；成立

blink [blɪŋk] v. 眨眼；闪烁；n. 眨眼；一瞬间

yawn [jɔːn] v. 打哈欠；张开；裂开；n. 哈欠

propagate [ˈprɒpəɡeɪt] v. 传播；宣传；繁殖；增殖

eliminate [ɪˈlɪmɪneɪt] v. 消除；排除；淘汰

modulation [ˌmɒdjʊˈleɪʃən] n. 调节；调整；调制

synthesis [ˈsɪnθəsɪs] n. 综合；结合；合成

superficial [ˌsuːpəˈfɪʃəl] adj. 表面的；肤浅的；浅薄的

perceive [pəˈsiːv] v. 察觉；感知；认为；理解

adaptively [əˈdæptɪvli] adv. 适应性地；有适应性地

enhance [ɪnˈhɑːns] v. 提高；增强；增进