Computers Math

Engineers at Tokyo Institute of Technology (Tokyo Tech) have demonstrated a simple computational approach for supporting the classification performance of neural networks operating on sensor time series. The proposed technique involves feeding the recorded signal as an external forcing into an elementary non-linear dynamical system, and providing its temporal responses to this disturbance to the neural network alongside the original data.
In the world around us, a proliferation of sensors is taking place, promising to support the efficiency and sustainability of practically all aspects of human activity. One challenge that engineers involved in delivering the internet-of-things to society have to face, is how to handle the flood of data resulting from such sensors. Especially, there is a need to reduce the data as much as possible at the edge, close to the sensors themselves, because streaming all data to the cloud would have an unacceptable technical, economic and environmental footprint. As a response to this, much research is being conducted worldwide towards small-sized, highly efficient classifiers suitable for detecting particular behaviors and situations of interest while running on limited computational resources. An example application scenario is the real-time monitoring of the behavior of livestock, having the purpose of detecting subtle changes that are indicative of prodromal disease.
“An emerging approach to support the development of time series classifiers suitable for edge artificial intelligence is that of data augmentation. Basically, it is about finding creative and innovative ways of generating additional data to help get the very best performance out of neural networks that necessarily have to be quite small to meet power and size requirements. While the theory of classifiers is well established, it can be said that data augmentation is still almost in its infancy for time series. In our laboratory, for example, we have been working on a variety of techniques based on empirical considerations as well as mathematical principles,” explains Ms. Chao Li, doctoral student at the Nano Sensing Unit where the study was conducted, and joint-lead author of the study.
Usually, data augmentation is performed just before or during classifier training, and runs on powerful workstations or cloud computers. The result is that the amount of data available to train a classifier is extended along the time dimension, as would be the case if longer recordings had been made available. This is important because high-quality data of the type necessary for classifier training is precious and expensive to prepare. However, this is not the only form of data augmentation possible. “We came up with the idea of extending the data along the other dimension, that is, the number of time series, meaning the number of input dimensions. Usually, edge applications may operate on one, or at most a few sensor time series. One possibility is performing computational operations to generate more of them, which try to make as much as possible of the initial information available to the classifier in a form suitable for it to learn it efficiently. While many signal processing operations could be implemented, a particularly disruptive computation is to simulate a dynamical system, endowed with its own intrinsic activity, and try to disturb it by externally forcing it with a signal recorded from the environment,” explains Dr. Ludovico Minati, lead author of the study.
Starting from a concept previously developed and patented in the Biointerfaces unit for improving the performance of brain-interface systems, the researchers carefully considered many practical aspects of how to realize it. Targeting the classification of the basic cattle behaviors using a collar-mounted accelerometer, they developed ways to filter and preprocess the kinematic signals and of injecting them so that the simulated dynamical system would accept and respond to them without diverging. Then, they explored how to extract the most relevant time series from its activity, in order to supply it either to a predetermined feature extractor and multi-layer perceptron or to a convolutional neural network. “Many low-dimensional systems such as the Rössler and Lorenz systems, which have been studied for decades by physicists and control engineers, actually have a remarkable computational potential that remains largely unexplored. This study takes an unusual step towards deploying it in a concrete application scenario,” explains Prof. Mattia Frasca from the University of Catania (Italy), who provided several theoretical contributions to the Tokyo Tech researchers on the behaviors of these kinds of systems and their implementations as analog circuits.
By augmenting the data through the additional time series derived from the dynamical systems, namely one separate Rössler system per accelerometer axis, the researchers were able to increase the classification performance by an appreciable amount. “While this is truly just an initial study to propose a provocative idea and substantial future work is needed, we were also able to realize the dynamical system using a very simple analog hardware circuit and still observe an improvement thanks to exploiting its responses,” adds Dr. Ludovico Minati. “Our approach reminds of reservoir computing, on which we recently conducted research using elementary transistor circuits known as the Minati-Frasca oscillators. However, it is actually different, because the dynamics are low dimensional, and a single oscillator is used instead of a network. In this sense, it may be even more suitable for low-power implementation” adds Mr. Jim Bartels, also a doctoral student at the unit.
After the interview, the team explained that this type of exploratory research will need to be extended and developed on other datasets and settings to ascertain its general applicability to concrete cases, though these initial results are promising. “One take-home point is that this approach can be implemented with quite limited resources, either digitally or in an analog way. Our past work in fact has shown CMOS chaotic systems operating with as low power as 1 μW, which could be suitable for this usage. As optimizations of process technologies and conventional designs approach their limits, the confident exploration of radically new ideas such as this one seems necessary for continued innovation,” concludes Dr. Hiroyuki Ito, head of the unit. The methodology, results and related considerations are reported in a recent article published in the journal Chaos, Solitons and Fractals, and all of the experimental recordings have been made freely available for others to use in future work. More

Mobile phone batteries with a lifetime up to three times longer than today’s technology could be a reality thanks to an innovation led by engineers at RMIT University.
Rather than disposing of batteries after two or three years, we could have recyclable batteries that last for up to nine years, the team says, by using high-frequency sound waves to remove rust that inhibits battery performance.
Only 10% of used handheld batteries, including for mobile phones, are collected for recycling in Australia, which is low by international standards. The remaining 90% of batteries go to landfill or are disposed of incorrectly, which causes considerable damage to the environment.
The high cost of recycling lithium and other materials from batteries is a major barrier to these items being reused, but the team’s innovation could help to address this challenge.
The team are working with a nanomaterial called MXene, a class of materials that they say promises to be an exciting alternative to lithium for batteries in the future.
Leslie Yeo, Distinguished Professor of Chemical Engineering and lead senior researcher, said MXene was similar to graphene with high electrical conductivity.

“Unlike graphene, MXenes are highly tailorable and open up a whole range of possible technological applications in the future,” said Yeo from RMIT’s School of Engineering.
The big challenge with using MXene was that it rusted easily, thereby inhibiting electrical conductivity and rendering it unusable, he said.
“To overcome this challenge, we discovered that sound waves at a certain frequency remove rust from MXene, restoring it to close to its original state,” Yeo said.
The team’s innovation could one day help to revitalise MXene batteries every few years, extending their lifetime up to three times, he said.
“The ability to prolong the shelf life of MXene is critical to ensuring its potential to be used for commercially viable electronic parts,” Yeo said.

The research is published in Nature Communications.
How the innovation works
Co-lead author Mr Hossein Alijani, a PhD candidate, said the greatest challenge with using MXene was the rust that forms on its surface in a humid environment or when suspended in watery solutions.
“Surface oxide, which is rust, is difficult to remove especially on this material, which is much, much thinner than a human hair,” said Alijani from RMIT’s School of Engineering.
“Current methods used to reduce oxidation rely on the chemical coating of the material, which limits the use of the MXene in its native form.
“In this work, we show that exposing an oxidised MXene film to high-frequency vibrations for just a minute removes the rust on the film. This simple procedure allows its electrical and electrochemical performance to be recovered.”
The potential applications of the team’s work
The team says their work to remove rust from Mxene opens the door for the nanomaterial to be used in a wide range of applications in energy storage, sensors, wireless transmission and environmental remediation.
Associate Professor Amgad Rezk, one of the lead senior researchers, said the ability to quickly restore oxidised materials to an almost pristine state represented a gamechanger in terms of the circular economy.
“Materials used in electronics, including batteries, generally suffer deterioration after two or three years of use due to rust forming,” said Rezk from RMIT’s School of Engineering.
“With our method, we can potentially extend the lifetime of battery components by up to three times.”
Next steps
While the innovation is promising, the team needs to work with industry to integrate its acoustics device into existing manufacturing systems and processes.
The team is also exploring the use of their invention to remove oxide layers from other materials for applications in sensing and renewable energy.
“We are keen to collaborate with industry partners so that our method of rust removal can be scaled up,” Yeo said. More

The diagnosis of Parkinson’s disease has shaken many lives. More than 10 million people worldwide are living with it. There is no cure, but if symptoms are noticed early, the disease can be controlled. As Parkinson’s disease progresses, along with other symptoms speech changes.
Lithuanian researcher from Kaunas University of Technology (KTU), Rytis Maskeliūnas, together with colleagues from the Lithuanian University of Health Sciences (LSMU), tried to identify early symptoms of Parkinson’s disease using voice data.
Parkinson’s disease is usually associated with loss of motor function — hand tremors, muscle stiffness, or balance problems. According to Maskeliūnas, a researcher at KTU’s Department of Multimedia Engineering, as motor activity decreases, so does the function of the vocal cords, diaphragm, and lungs: “Changes in speech often occur even earlier than motor function disorders, which is why the altered speech might be the first sign of the disease.”
Expanding the AI language database
According to Professor Virgilijus Ulozas, at the Department of Ear, Nose, and Throat at the LSMU Faculty of Medicine, patients with early-stage of Parkinson’s disease, might speak in a quieter manner, which can also be monotonous, less expressive, slower, and more fragmented, and this is very difficult to notice by ear. As the disease progresses, hoarseness, stuttering, slurred pronunciation of words, and loss of pauses between words can become more apparent.
Taking these symptoms into account, a joint team of Lithuanian researchers has developed a system to detect the disease earlier.

“We are not creating a substitute for a routine examination of the patient — our method is designed to facilitate early diagnosis of the disease and to track the effectiveness of treatment,” says KTU researcher Maskeliūnas.
According to him, the link between Parkinson’s disease and speech abnormalities is not new to the world of digital signal analysis — it has been known and researched since the 1960s. However, as technology advances, it is becoming possible to extract more information from speech.
In their study, the researchers used artificial intelligence (AI) to analyse and assess speech signals, where calculations are done and diagnoses made in seconds rather than hours. This study is also unique — the results are tailored to the specifics of the Lithuanian language, in this way expanding the AI language database.
The algorithm will become a mobile app in the future
Speaking about the progress of the study, Kipras Pribuišis, lecturer at the Department of Ear, Nose, and Throat at the LSMU Faculty of Medicine, emphasises that it was only carried out on patients already diagnosed with Parkinson’s: “So far, our approach is able to distinguish Parkinson’s from healthy people using a speech sample. This algorithm is also more accurate than previously proposed.”
In a soundproof booth, a microphone was used to record the speech of healthy and Parkinson’s patients, and an artificial intelligence algorithm “learned” to perform signal processing by evaluating these recordings. The researchers highlight that the algorithm does not require powerful hardware and could be transferred to a mobile app in the future.
“Our results, which have already been published, have a very high scientific potential. Sure, there is still a long and challenging way to go before it can be applied in everyday clinical practice,” says Maskeliūnas.
According to the researcher, the next steps include increasing the number of patients to gather more data and determining whether the proposed algorithm is superior to alternative methods used for early diagnosis of Parkinson’s. In addition, it will be necessary to check whether the algorithm works well not only in laboratory-like environments but also in the doctor’s office or in the patient’s home. More