Abstract: A voice recognition method of an artificial intelligence robot device is disclosed. The voice recognition method includes collecting a first voice spoken by a user and determining whether a wake-up word of the artificial intelligence robot device is recognized based on the collected first voice; if the wake-up word is not recognized, sensing a location of the user using at least one sensor and determining whether the sensed location of the user is included in a set voice collection range; if the location of the user is included in the voice collection range, learning the first voice and determining a noise state of the first voice based on the learned first voice; collecting a second voice in an opposite direction of the location of the user according to a result of the determined noise state of the first voice; and extracting a feature value of a noise based on the second voice and removing the extracted feature value of the noise from the first voice to obtain the wake-up word.

Abstract: An electronic device is provided. The electronic device includes a microphone to receive audio, a communicator, a memory configured to store computer-executable instructions, and a processor configured to execute the computer-executable instructions. The processor is configured to determine whether the received audio includes a predetermined trigger word; based on determining that the predetermined trigger word is included in the received audio; activate a speech recognition function of the electronic device; detect a movement of a user while the speech recognition function is activated; and based on detecting the movement of the user, transmit a control signal, to a second electronic device to activate a speech recognition function of the second electronic device.

Abstract: Techniques described herein use backpropagation to train one or more machine learning (ML) models of a dialog system. For instance, a method includes accessing seed data that includes training tuples, where each training tuple comprising a respective logical form. The method includes converting the logical form of a training tuple to a converted logical form, by applying to the logical form a text-to-speech (TTS) subsystem, an automatic speech recognition (ASR) subsystem, and a semantic parser of a dialog system. The method includes determining a training signal by using an objective function to compare the converted logical form to the logical form. The method further includes training the TTS subsystem, the ASR subsystem, and the semantic parser via backpropagation based on the training signal. As a result of the training by backpropagation, the machine learning models are tuned work effectively together within a pipeline of the dialog system.

Abstract: Disclosed is an approach to implement improved anomaly detection. Improved anomaly detection is provided using MSET-SPRT via Monte Carlo simulation that can address problems with conventional MSET-SPRT approaches and provide improved system performance and accuracy.

Abstract: The disclosure provides technology for enabling a computing device to provide context sensitive special effects that supplement a text source as it is read aloud. An example method includes receiving, by a processing device, audio data comprising a spoken word of a user, analyzing contextual data associated with the user, determining a match between the audio data and data of a text source; and initiating a physical effect in response to the determining the match, wherein the physical effect corresponds to the text source and is based on the contextual data.

Abstract: In industrial machine predictive maintenance system may include an industrial machine data analysis facility that generates streams of industrial machine health monitoring data by applying machine learning to data representative of conditions of portions of industrial machines received via a data collection network. The system may perform a method of sampling a signal at a streaming sample rate to produce a plurality of samples of the signal. Portions of the plurality of samples can be allocated to first and second signal analysis circuits based on signal analysis sampling rates less than the streaming sample rate, and the samples and the outputs of the signal analysis circuits can be stored. The system can include a sensor detecting a condition of an industrial machine to output a signal, which can be sampled at a streaming sample rate that is at least twice a dominant frequency of the signal.

Abstract: Example implementations involve a system for Predictive Maintenance using Generative Adversarial Networks for Failure Prediction. Through utilizing three processes concurrently and training them iteratively with data-label pairs, example implementations described herein can thereby generate a more accurate predictive maintenance model than that of the related art. Example implementations further involve shared networks so that the three processes can be trained concurrently while sharing parameters with each other.

Abstract: A biological information detection apparatus includes: a light source which projects a pattern of near-infrared light onto an object including a living body; an imaging system which includes first photodetector cells detecting light in a near-infrared wavelength range and second photodetector cells detecting light in a visible wavelength range, and generates a first image signal representing a first image, which is an image taken in the near-infrared wavelength range, of the object on which the pattern is projected, and a second image signal representing a second image of the object taken in the visible wavelength range; and a calculator which calculates biological information concerning the living body using at least one selected from the group consisting of the first and second image signals.

Abstract: According to one embodiment, an information processing method includes performing, in an intermediate layer of a deep neural network, a forward propagation using a first parameter and based on a first input value represented by a first bit number; performing quantization to produce a second input value represented by a second bit number smaller than the first bit number, and storing the produced second input value in the memory; calculating a second parameter based on a result of an operation using the second input value stored in the memory and a value obtained by the forward propagation, the second parameter being an update of the first parameter and for use in the learning process; and determining a condition for the quantization based on a gradient difference obtained in said calculating the second parameter.

Abstract: A measurement device and a measurement method capable of measuring a depth of a damage source in a structure having a thickness of a predetermined value or more. According to an embodiment, a measurement device includes a first detector and a signal processing device. The first detector selectively detects surface waves that are excited when first elastic waves generated inside a structure formed of a solid material have reached a surface of the structure. The information processing device obtains information about a depth of a source of the first elastic waves within the structure on the basis of information of at least one of an amplitude and a time of arrival of the surface waves detected by the first detector.

Abstract: This disclosure relates to a method and system for monitoring health and predicting failure of an electro-mechanical machine. In an embodiment, the method may include receiving a plurality of operational parameters with respect to the electro-mechanical machine and determining a set of features and a set of events, based on the plurality of operational parameters. The method may further include detecting one or more fault signatures associated the electro-mechanical machine based on at least one of the plurality of operational parameters, the set of features, or the set of events. The method may further include determining at least one of a time to the possible failure and a remaining useful life of the electro-mechanical machine based on at least one of the plurality of operational parameters, the set of features, the set of events, or the one or more fault signature, by using a hybrid machine learning model.

Abstract: A computer system is configured to group solar power systems that provide electric power to an electricity distribution system into clusters. The computer system identifies a solar source meter in each of the clusters that is representative of the respective one of the clusters as a bellwether meter. Each of the bellwether meters monitors a power output of one of the solar power systems in one of the clusters. The computer system receives solar power generation data from the bellwether meters. The computer system generates a solar power generation forecast for each of the clusters of the solar power systems using the solar power generation data from the bellwether meters in respective ones of the clusters.

Abstract: A method for generating forecast predictions that indicate an event horizon of an entity or remaining useful life of a consumable using machine learning techniques is provided. Using a server computer system, feature data comprising features vectors that represent a set of signal data over a range of time is stored. Condition data comprising conditions occurring on the entity at particular moments in time is stored. Label data that comprises a plurality of time values that each indicate a difference in time between one condition and another condition is stored. A training dataset is created by combining the feature data, the condition data, and the label data into a single dataset. The training dataset is partitioned by condition. A machine learning model is trained on each target condition training dataset. The trained machine learning models are used to generate forecast values that each indicate an amount of time to an occurrence of a target condition associated with an entity.

Abstract: Audio systems, methods, and processor instructions are provided that detect voice activity of a user and provide an output voice signal. The systems, methods, and instructions receive a plurality of microphone signals and combine the plurality of microphone signals according to a first combination and a second combination. The first combination produces a primary signal having enhanced response in the direction of the user's mouth, and the second combination produces a reference signal having reduced response in the direction of the user's mouth. The primary signal and the reference signal are added and subtracted to produce a summation signal and a difference signal, respectively. The summation signal and the difference signal are compares and an output voice signal is provided based upon the comparison.

Abstract: Systems and methods for processing measurement data in an electric power system include acquiring the measurement data by a phasor measurement unit (PMU) coupled to a line of the electric power system, and inputting a plurality of the measurement data within a predetermined time window into a K-nearest neighbor (KNN) for identifying bad data among the plurality of the measurement data, wherein when one of the plurality of measurement data contains a bad datum, the machine learning module sends the bad datum to a denoising autoencoder module for correcting the bad datum, wherein the denoising autoencoder module outputs a corrected part corresponding to the bad datum, and when one of the plurality of measurement data contains no bad datum, the machine learning module bypasses the denoising autoencoder module and outputs the one of the plurality of measurement data as an untouched part.

Abstract: Described herein are embodiments of an end-to-end text-to-speech (TTS) system with parallel wave generation. In one or more embodiments, a Gaussian inverse autoregressive flow is distilled from an autoregressive WaveNet by minimizing a novel regularized Kullback-Leibler (KL) divergence between their highly-peaked output distributions. Embodiments of the methodology computes the KL divergence in a closed-form, which simplifies the training process and provides very efficient distillation. Embodiments of a novel text-to-wave neural architecture for speech synthesis are also described, which are fully convolutional and enable fast end-to-end training from scratch. These embodiments significantly outperform the previous pipeline that connects a text-to-spectrogram model to a separately trained WaveNet. Also, a parallel waveform synthesizer embodiment conditioned on the hidden representation in an embodiment of this end-to-end model were successfully distilled.

Abstract: Techniques enable an automatic adjustment of a muted response setting of an automated assistant based on a determination of an expectation by a user to hear an audible response to their query, despite the muted setting. Determination of the expectation may be based on historical, empirical data uploaded from multiple users over time for a given response scenario. For example, the system may determine from the historical data that a certain type of query has been associated with a user both repeating their query and increasing a response volume setting within a given timeframe. Metrics may be generated, stored, and invoked in response to attributes associated with identifiable types of queries and query scenarios. Automated response characteristics meant to reduce inefficiencies may be associated with certain queries that can otherwise collectively burden network bandwidth and processing resources.

Abstract: Disclosed are an automatic question answering method and apparatus, a storage medium, and a server. The method includes: acquiring numerical features of a sentence to be queried; querying a target sentence in a question database according to the numerical features of the sentence to be queried, the question database including a plurality of sentences and answers corresponding to the plurality of sentences; and determining a target answer according to an answer corresponding to the target sentence. In this method, the sentence is represented by the numerical features, such that it is convenient to search questions similar to a question of a user in the question database, thereby achieving an effect of improving a search speed of the question.

Abstract: Methods and systems for failure prediction using analysis of oil or other lubricant. Raw data about feature(s) of each of a plurality of particles filtered from a fluid sample are used to categorize each particle into one of a plurality of categories, each category being defined by one or more of: chemical composition, size and morphology. Particle physical characteristics in each category are quantified to obtain a set of categorized data. The categorized data are compared with historical data. Results of the comparing are evaluated to generate a prediction of any failure or mechanism of failure.

Abstract: Disclosed is an electronic apparatus providing a reply to a query of a user. The electronic apparatus includes a microphone, a camera, a memory configured to store at least one instruction, and at least one processor, and the processor is configured to execute the at least one instruction to control the electronic apparatus to: identify a region of interest corresponding to a co-reference in an image acquired through the camera based on a co-reference being included in the query, identify an object referred to by the co-reference among at least one object included in the identified region of interest based on a dialogue content that includes the query, and provide information on the identified object as the reply.