System Bits: July 31

Computers that perceive human emotion
As part of the growing field of “affective computing,” MIT researchers have developed a machine-learning model that takes computers a step closer to interpreting our emotions as naturally as humans do.

Affective computing uses robots and computers to analyze facial expressions, interpret emotions, and respond accordingly. Applications include, for instance, monitoring an individual’s health and well-being, gauging student interest in classrooms, helping diagnose signs of certain diseases, and developing helpful robot companions.

MIT Media Lab researchers have developed a machine-learning model that takes computers a step closer to interpreting our emotions as naturally as humans do. The model better captures subtle facial expression variations to better gauge moods. By using extra training data, the model can also be adapted to an entirely new group of people, with the same efficacy.
Source: MIT

The researchers pointed out one challenge to this is that people express emotions quite differently, depending on many factors, but general differences can be seen among cultures, genders, and age groups. Other differences, however, are even more fine-grained such as the time of day, how much you slept, or even your level of familiarity with a conversation partner leads to subtle variations in the way you express, say, happiness or sadness in a given moment.

They noted that human brains instinctively catch these deviations, but machines struggle, and even though deep-learning techniques have been developed in recent years to help catch the subtleties, they’re still not as accurate or as adaptable across different populations as they could be.
As such, the team developed a machine-learning model that they say can outperform traditional systems in capturing these small facial expression variations to better gauge mood while training on thousands of images of faces. Further, they noted that by using a little extra training data, the model can be adapted to an entirely new group of people, with the same efficacy, in order to improve existing affective-computing technologies.

To learn more about how they accomplished this, click here.

One promising application, the researchers pointed out, is human-robotic interactions, such as for personal robotics or robots used for educational purposes, where the robots need to adapt to assess the emotional states of many different people. One version, for instance, has been used in helping robots better interpret the moods of children with autism.

Light-emitting nanocrystals
In a development that brings plasmonics research a step closer to realizing ultra-compact light sources for high-speed, optical data processing and other on-chip applications, University of California San Diego researchers have built a nanosized device out of silver crystals using advanced fabrication techniques that can generate light by efficiently “tunneling” electrons through a tiny barrier.

Illustration of nanosized device made of two joined silver single crystals that generate light by inelastical electron tunneling.
Source: UCSD

The team explained that the device emits light by a quantum mechanical phenomenon known as inelastic electron tunneling, in which electrons move through a solid barrier that they cannot classically cross. And even though the electrons lose some of their energy while crossing, they create either photons or phonons in the process.

Plasmonics researchers have been interested in using inelastic electron tunneling to create extremely small light sources with large modulation bandwidth but since only a tiny fraction of electrons can tunnel inelastically, the efficiency of light emission is typically low—on the order of a few hundredths of a percent, at most, the UCSD team said.

To address this, the researchers created a device that bumps that efficiency up to approximately 2 percent, and while this is not yet high enough for practical use, it is the first step to a new type of light source.

This light emitting device was designed using computational methods and numerical simulations, and constructed using advanced solution-based chemistry techniques.

Left: schematics of the tunnel junction formed by two edge-to-edge silver single crystal cuboids with an insulating barrier of polyvinylpyrrolidone (PVP). The top inset shows that photons are generated through inelastic electron tunneling. The device performance can be engineered by tuning the size of the cuboids (a, b, c), the gap size (d), and the curvature of silver cuboid edges. Right: TEM image of the tunnel junction, where the gap is around 1.5 nm.
Source: UCSD

With additional work, the team said they aim to further boost efficiency another order of magnitude higher, and are exploring different geometries and materials for future studies.

Excitons take electronics to future
In a breakthrough that could lead to a new breed of faster, more energy efficient and smaller electronics, EPFL researchers have developed a transistor based on excitons – a type of particle most people have not heard of – that is able to function at room temperature..

The team said excitons could revolutionize the way engineers approach electronics. And this new type of transistor uses these particles instead of electrons.

Interestingly, their exciton-based transistor functions effectively at room temperature, which has previously been an insurmountable obstacle. The researchers said they achieved this by using two 2D materials as semiconductors.

They believe this sets the stage for optoelectronic devices that consume less energy and are both smaller and faster than current devices. In addition, it will be possible to integrate optical transmission and electronic data-processing systems into the same device, which will reduce the number of operations needed and make the systems more efficient.