Speech activity detection (SAD) plays an important role in current speech processing systems, including automatic speech recognition (ASR). SAD is particularly difficult in environments with acoustic noise. A practical solution is to incorporate visual information, increasing the robustness of the SAD approach. An audiovisual system has the advantage of being robust to different speech modes (e.g., whisper speech) or background noise. Recent advances in audiovisual speech processing using deep learning have opened opportunities to capture in a principled way the temporal relationships between acoustic and visual features. This study explores this idea proposing a \emph{bimodal recurrent neural network} (BRNN) framework for SAD. The approach models the temporal dynamic of the sequential audiovisual data, improving the accuracy and robustness of the proposed SAD system. Instead of estimating hand-crafted features, the study investigates an end-to-end training approach, where acoustic and visual features are directly learned from the raw data during training. The experimental evaluation considers a large audiovisual corpus with over 60.8 hours of recordings, collected from 105 speakers. The results demonstrate that the proposed framework leads to absolute improvements up to 1.2% under practical scenarios over a VAD baseline using only audio implemented with deep neural network (DNN). The proposed approach achieves 92.7% F1-score when it is evaluated using the sensors from a portable tablet under noisy acoustic environment, which is only 1.0% lower than the performance obtained under ideal conditions (e.g., clean speech obtained with a high definition camera and a close-talking microphone).

A good neural sequence-to-sequence summarization model should have a strong encoder that can distill and memorize the important information from long input texts so that the decoder can generate salient summaries based on the encoder's memory. In this paper, we aim to improve the memorization capabilities of the encoder of a pointer-generator model by adding an additional 'closed-book' decoder without attention and pointer mechanisms. Such a decoder forces the encoder to be more selective in the information encoded in its memory state because the decoder can't rely on the extra information provided by the attention and possibly copy modules, and hence improves the entire model. On the CNN/Daily Mail dataset, our 2-decoder model outperforms the baseline significantly in terms of ROUGE and METEOR metrics, for both cross-entropy and reinforced setups (and on human evaluation). Moreover, our model also achieves higher scores in a test-only DUC-2002 generalizability setup. We further present a memory ability test, two saliency metrics, as well as several sanity-check ablations (based on fixed-encoder, gradient-flow cut, and model capacity) to prove that the encoder of our 2-decoder model does in fact learn stronger memory representations than the baseline encoder.

Abstract-- We investigate a reinforcement approach for distributed sensing based on the latent space derived from multimodal deep generative models. Our contribution provides insights to the following benefits: Detections can be exchanged effectively between robots equipped with unimodal sensors due to a shared latent representation of information that is trained by a Variational Auto Encoder (VAE). Sensor-fusion can be applied asynchronously due to the generative feature of the VAE. Deep Q-Networks (DQNs) are trained to minimize uncertainty in latent space by coordinating robots to a Point-of- Interest (PoI) where their sensor modality can provide beneficial information about the PoI. Additionally, we show that the decrease in uncertainty can be defined as the direct reward signal for training the DQN.

The El Niño/La Niña pattern in the Pacific Ocean is notorious for its long-distance effects on weather as far away as Africa and the Midwestern United States. But climate experts also know of several other such patterns, known as "teleconnections," and believe that there are many more to be discovered. The new TRIPODS Climate project, a collaboration among the University of Chicago, University of Wisconsin-Madison and the University of California-Irvine, will develop novel data science tools to sniff out these hidden patterns, improving weather forecasts and scientific understanding of global climate. Researchers will apply data science methods such as machine learning, network analysis and predictive modeling to the growing flood of climate data. "There are fundamental challenges pervasive in data science that are epitomized in the climate science setting, making this collaboration a nice opportunity for advances on a number of fronts," said Rebecca Willett, professor of computer science and statistics at UChicago.

Soul Machines has discussed services for the elderly with prospective clients but has not announced any partnerships on that subject to date, says chief business officer Greg Cross. Soul Machines envisions a future in which digital instructors educate students without access to quality human teachers, and in which famous deceased artists are digitally resurrected to discuss their works in museums. Robot companions for the infirm, then, are not too far a leap. Nor is the prospect of a future in which a family converses with the lively AI recreation of a person suffering from dementia, while a caregiver--robot or human--tends to their ailing body in another room. The potential for deception is already here. A few years ago, Brent Lawson, the president of 1 AM Dolls, a manufacturer of life-sized rubber sex dolls, was on the phone with a client who wanted a specific doll he'd seen on the company's website. The man was particularly concerned that the doll's hair was just so, and peppered Lawson with questions about the color and style, Lawson told Quartz.

Artificial Intelligence (AI) is changing how we commute and experience urban living. As bike share schemes grow, AI is playing an increasing role in successful schemes and enabling operators to grow efficiently in new markets. While some bike share schemes are struggling to maximise profitability and maintain market share due to increasing headcount and throwing expensive resources into their schemes as they grow, others are using new technologies to accelerate ridership growth and optimise their operations. To succeed in the long-term, they have had to rethink their approach to bike share scheme management and find ways to use the data in their schemes and turn it into actionable insights. Many operators have global ambitions and want to expand into new markets but the first step is to ensure they have the best possible operating model and can replicate it in new cities.

Toddlers use the same gestures as chimpanzees and gorillas showing they really are just'tiny apes', claim researchers. One to two year olds use 52 limb and body movements to communicate - nine in ten of which are observed in great apes. This is a crucial stage of development when infants are on the cusp of learning language, say Scottish scientists behind the findings. Toddlers use the same gestures as chimpanzees and gorillas showing they really are just'tiny apes', claim researchers. Senior author Dr Catherine Hobaiter, of the School of Psychology and Neuroscience at St Andrews University, said: 'Wild chimpanzees, gorillas, bonobos and orangutans all use gestures to communicate their day-to-day requests.

Hurricane Harvey, shown in 2017. A new data project hopes to sniff out weather patterns. The El Nino and La Nina patterns in the Pacific Ocean are notorious for their long-distance effects on weather as far away as Africa and the Midwestern United States. But climate experts also know of several other such patterns, known as teleconnections, and believe that there are many more to be discovered. The new TRIPODS Climate project, a collaboration among the University of Wisconsin–Madison, the University of Chicago, and the University of California, Irvine, will develop novel data science tools to sniff out these hidden patterns, improving weather forecasts and scientific understanding of global climate.

Microsoft's new Azure edge computing offerings are helping customers extend the reach of its cloud-based machine learning services, according to Clayton Fernandez, the company's global director, Internet of Things. In June this year, Azure IoT Edge hit general availability, offering customers of Microsoft's cloud service new capabilities including support for the Moby container management system at the edge, the Azure IoT Edge security manager, an Automatic Device Management (ADM) service, and a range of tools for developers. Coinciding with the announcement that it had been GAed, Microsoft open sourced the IoT Edge code and posted it on GitHub. "We started with putting a PC on everyone's desk," Fernandez told Computerworld. "Next came everyone having smartphones in their pockets -- but now things are going to start getting really interesting with this whole universe of interconnected devices that are coming together in the intelligent cloud. "Today we're accepting billions of signals securely; we ingest, we reason, we take action – we call it the intelligent edge because these devices are becoming so capable that they're helping power some of the most advanced algorithms.

AUT's Artificial Intelligence knows your choices before you do Picture standing in your local dairy, indecisively scanning the drinks fridge to choose which can you're going to buy. If you were hooked up to NeuCube, it would be able to predict which one you'd reach for – before you'd opened the fridge, and even before the thought was fully formed. Using artificial intelligence to predict a person's choices before they have even made up their mind is a world first - and it was developed at Auckland University of Technology (AUT). Professor Nikola Kasabov, who worked on the project, said they wanted to find out if there was brain activity before the conscious perception of an object. READ MORE: * Artificial intelligence presents us with an opportunity, and a challenge * Don't fear the new AI revolution * How computers'see' faces and other objects * How science may just end up killing crime There is, it turns out.