Abstract: Example methods disclosed herein to initialize classification vectors of a neural network include ranking a plurality of classes to be represented by the classification vectors based on respective numbers of instances of training data associated with corresponding ones of the classes. Disclosed example methods also include initializing the classification vectors to span a classification space corresponding to the classes. Disclosed example methods further include assigning respective ones of the classes to corresponding ones of the classification vectors based on the ranking of the classes.

Abstract: An audio metadata providing apparatus and method and a multichannel audio data playback apparatus and method to support a dynamic format conversion are provided. Dynamic format conversion information may include information about a plurality of format conversion schemes that are used to convert a first format set by a writer of multichannel audio data into a second format that is based on a playback environment of the multichannel audio data and that are set for each of playback periods of the multichannel audio data. The audio metadata providing apparatus may provide audio metadata including the dynamic format conversion information. The multichannel audio data playback apparatus may identify the dynamic format conversion information from the audio metadata, may convert the first format of the multichannel audio data into the second format based on the identified dynamic format conversion information, and may play back the multichannel audio data with the second format.

Abstract: An electronic device for active learning includes at least one memory and at least one processor coupled to the at least one memory. The at least one processor is configured to select one or more entries from a data set including unlabeled data based on a similarity between the one or more entries and labeled data. The at least one processor is further configured to cause the one or more entries to be labeled.

Abstract: A method and system are provided for extracting features from digital audio signals which exhibit variations in pitch, timbre, decay, reverberation, and other psychoacoustic attributes and learning, from the extracted features, an artificial neural network model for generating contextual latent-space representations of digital audio signals. A method and system are also provided for learning an artificial neural network model for generating consistent latent-space representations of digital audio signals in which the generated latent-space representations are comparable for the purposes of determining psychoacoustic similarity between digital audio signals.

Abstract: Embodiments are related to processing of one or more input audio feeds for generation of a target audio stream that includes at least one object of interest to a listener. In some embodiments, the target audio stream may exclusively or primarily include the sound of the object of interest to the listener, without including other persons. This allows a listener to focus on an object of his or her interest and not necessarily have to listen to the performances of other objects in the input audio feed. Some embodiments contemplate multiple audio feeds and/or with multiple objects of interest.

Abstract: A voice processing system includes a voice receiver that receives voice spoken by users, a voice recognition unit that recognizes the voice received by the voice receiver, a voice determination unit that determines a state of transmission and reception of the voice between a first base and a second base based on a recognition result of the first voice received from the first base and a recognition result of the second voice received from the second base, and an information presenting unit that presents specific information for resolving an abnormal state of transmission and reception when it is determined by the voice determination unit that the state of transmission and reception is abnormal.

Abstract: A method of electronically documenting a conversation is provided. The method includes capturing audio of a conversation between a first speaker and a second speaker; generating conversation audio data from the captured audio; and segmenting the conversation audio data into a plurality of utterances according to a speaker segmentation technique. The method further includes, for each utterance: storing time data indicating the chronological position of the utterance in the conversation; passing the utterance to a neural network model, the neural network model configured to receive the utterance as an input and generate a feature representation of the utterance as an output; assigning the utterance feature representation to a first speaker cluster or a second speaker cluster according to a clustering technique; assigning a speaker identifier to the utterance based on the cluster assignment of the utterance; and generating a text representation of the utterance.

Abstract: The present disclosure relates to the field of medical data processing based on natural language processing. Embodiments of the present disclosure disclose a method and apparatus for verifying a medical fact. The method may include: acquiring a description text of the medical fact; selecting a relevant paragraph related to the description text of the medical fact from a medical document; and inputting the description text of the medical fact and the corresponding relevant paragraph into a trained discrimination model for authenticity judgment, to obtain a verification result of the medical fact, the discrimination model being pre-trained based on a medical text paragraph pair extracted from the medical document, and being iteratively adjusted using a medical fact sample set including authenticity labeling information after the pre-training.

Abstract: The present disclosure provides a method for controlling data, a device for controlling data, an electronic equipment, and a storage medium. The method for controlling data includes receiving a first voice signal for a virtual object displayed on a first electronic device collected by the first electronic device; obtaining a first recognition result by parsing and recognizing the first voice signal; generating, based on the first recognition result, a control instruction for controlling the virtual object to perform an action corresponding to the control instruction; and sending the control instruction to the first electronic device and at least one second electronic device in the same virtual space. The disclosure is capable of controlling a process of the virtual object performing a live streaming interaction by using a voice interaction mode, thus reducing user operations and improving user experience.

Abstract: The invention relates to a portable device (1100) comprising a bottom surface (BS) intended to be in contact with a textile (TXT). The portable device comprises an image sensor (5) for taking an image of a textile, and an illumination system (6) for illuminating a portion of the textile when said image is being taken. The portable device also comprises a control unit (8a) for executing an algorithm stored in the portable device, using the taken image as an input of said algorithm, to obtain a classification of the textile. The image sensor and the control unit are integrated within the portable device. The image sensor has an active surface sensitive to light, which is oriented compared to said bottom surface (BS), with an absolute value of the orientation angle being in the range 15-70 degrees; and/or the illumination system (6) has a light source oriented compared to the said bottom surface (BS), with an absolute value of the orientation angle being in the range 15-70 degrees.

Abstract: At least some embodiments are directed to an entity classification system receives informational data associated with an entity. The informational data includes sentences associated with the entity. The entity classification system utilizes a first machine learning model to determine a first contextual meaning among words of a sentence associated with the entity based on a first word embedding technique, and determines at least one category associated with the entity based at least in part on the first contextual meaning. The entity classification system utilizes a second machine learning model to determine a second contextual meaning shared by a set of sentences based on a second embedding technique, and determines a subcategory of the category associated with the entity based at least in part on the second contextual meaning. The entity classification system generates an output including the category and subcategory associated with the entity.

Abstract: System and methods for processing audio signals are disclosed. In one implementation, a system may comprise a wearable camera configured to capture images from an environment of a user; a microphone; and a processor. The processor may be configured to receive an audio signal representative of sounds captured by the microphone during a time period; and receive the images captured by the wearable camera. The processor may process the audio signal in a first mode based on audio data accumulated in a buffer prior to the time period; detect a change in the active speaker from the first individual to a second individual; and cease processing in the first mode and process the audio signal in a second mode that differs from the first mode.

Abstract: A virtual space view is provided in an online meeting user interface associated with a videoconferencing system that is changeable by gestures of meeting participants. The virtual space view displays avatars or video feeds of the meeting participants on their respective participant computers. The videoconferencing system captures video feeds of all meeting participants, including meeting participants being displayed as avatars. Software of the videoconferencing system recognizes a specific gesture of a meeting participant. The specific gesture is associated with a predefined directional movement of the meeting participant's avatar or video feed within the virtual space view during the online meeting, and the predefined directional movement is requested by the meeting participant via the specific gesture. The specific gesture of the meeting participant is a physical body gesture of the meeting participant which is captured in the video feed of the meeting participant.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for selecting a neural network architecture for performing a machine learning task. In one aspect, a method comprises: obtaining data defining a synaptic connectivity graph representing synaptic connectivity between neurons in a brain of a biological organism; generating data defining a plurality of candidate graphs based on the synaptic connectivity graph; determining, for each candidate graph, a performance measure on a machine learning task of a neural network having a neural network architecture that is specified by the candidate graph; and selecting a final neural network architecture for performing the machine learning task based on the performance measures.

Abstract: A system or method for dynamic redundant call recording may include a plurality of recording devices, each recording device having a plurality of recording resources, and a resource allocator. The resource allocator may receive a request from a call receiving node for commencement of a recording session. It may then attempt to connect to a first one of the plurality of recording devices and if successful, establish a recording session between the call receiving node and the recording device, or if not successful, attempting to connect the recording session controller to a second one of the multiple recording devices. Two resource allocators may operate in parallel to establish dual recording using resources at two different recording devices. Call content may be recorded separately from call metadata and may be integrated with the metadata using a unique call ID.

Abstract: Provided is a portable telephone or an earphone or an electromagnetic-type vibration unit that comprises an electromagnetic-type vibrator including two mutually relatively moving portions supported via an elastic body, and that is configured to produce vibrations from at least one of the two portions for cartilage conduction. The elastic body provides a cartilage conduction portion to be contacted with the ear cartilage. The elastic body is disposed on or in the vicinity of an upper corner portion of a housing. One of the two portions is supported on a housing structure with the elastic body supported by the housing structure, and the other of the two portions is supported by the elastic body. Of the two portions, one with a greater weight is supported by the elastic body. Of the two portions, one or a portion connected thereto provides a cartilage conduction portion to be contacted with the ear cartilage, with the elastic body disposed between the two portions.

Abstract: A camera is auto-focused on an in-person attendee in a hybrid meeting who is speaking into a microphone at a live location of the hybrid meeting, wherein the hybrid meeting is being streamed via a videoconferencing system to one or more remote attendees. The live location includes a real-time location system (RTLS) having a RTLS tag and a plurality of RTLS anchors for locating a position of the RTLS tag in three-dimensional space. The RTLS tag is fixed to the microphone. In operation, it detected when the microphone is active. The position of the RTLS tag in the live location is then identified, wherein the position of the RTLS tag in the live location is presumed to be the same position as the microphone that the RTLS tag is fixed to.

Abstract: A system and method allows for access to the contacts of one person on a second person's device, without affecting the content on the second person's device. The contact information, and other information that is stored on a first person's device, can be temporarily downloaded onto a second person's device for making calls or accessing the other information, thereby allowing for contact with someone. Upon logging out of the system on the second device, the information from the first person is deleted. This allows the first person to contact others when the first person's phone is not available for some reason. The system also allows for the downloading of more content and virtual access to the lost/stolen telephone to be able to locate the phone, control the phone or destroy the information on the phone.

Abstract: A mirrored gallery view is provided of a breakout room in an online meeting user interface associated with a videoconferencing session with a session view established in a videoconferencing system. The mirrored gallery view displays video feeds of meeting participants on their respective participant computers. The video feeds are camera-captured views of each of the meeting participants, The videoconferencing system creates a breakout room within the videoconferencing session for a subset of the meeting participants, thereby allowing the subset of the meeting participants to engage with one another within the breakout room during the videoconferencing session. A video processor automatically creates mirrored views of the video feed of each of the subset of meeting participants in the breakout room whose video feed in the videoconferencing session is not currently mirrored.

Abstract: The present invention relates to coding of audio signals, and in particular to high frequency reconstruction methods including a frequency domain harmonic transposer. A system and method for generating a high frequency component of a signal from a low frequency component of the signal is described.