Abstract: A computer-implemented method for generating a custom thumbnail is disclosed. The computer-implemented method includes segmenting an audio file into one or more audio segments based, at least in part, on a respective context associated with each of the one or more audio segments. The computer-implemented method further includes selecting an initial thumbnail image based, at least in part, on one or more contexts associated with the one or more audio segments. The computer-implemented method further includes generating a customized thumbnail image based, at least in part, on performing a neural style transfer of the initial thumbnail image and a style reference image.

Abstract: [Problem] To provide an audio-input device with which it is possible to properly acquire also the audio of a conversation held between a wearer of said device and two other conversation participants.[Solution] This audio-input device 100 is provided with: a first arm 10 and a second arm 20 which can be arranged at positions such that a target sound source is interposed therebetween; and a plurality of sound collection parts 41-46 provided in a quantity of at least three for each of the first arm 10 and the second arm 20 (a total quantity of at least six).

Abstract: The present disclosure relates to a method and system for performing packet loss concealment using a neural network system. The method comprises obtaining a representation of an incomplete audio signal, inputting the representation of the incomplete audio signal to an encoder neural network and outputting a latent representation of a predicted complete audio signal. The latent representation is input to a decoder neural network which outputs a representation of a predicted complete audio signal comprising a reconstruction of the original portion of the complete audio signal, wherein said encoder neural network and said decoder neural network have been trained with an adversarial neural network.

Abstract: A system and method of real-time feedback confirmation to solicit a virtual assistant response from an evolving semantic state of at least a portion of an utterance. A user accesses a virtual assistant on an electronic device having the system and/or method configured to capture a command, a question, and/or a fulfillment request from audio such as, the speech emitted from the speaking user. The speech may be intercepted by a speech engine configured to transcribe the speech into text that is matched with the fragment pattern's regular expression to generate a fragment and/or the speech may be processed with a machine learning model to identify fragments. The fragments are identified by a domain handler configured to update a data structure of the current semantic state of the utterance in real-time on an interface of an electronic device.

Abstract: Systems and methods for detecting and tracking conversations of interest in crowded areas. The system includes a plurality of body worn portable public safety communication devices, each including a plurality of microphones and an electronic processor. The electronic processor is configured to generate, using the plurality of microphones, a plurality of beams and scan in parallel, using the plurality of beams, a plurality of simultaneous audio conversations. The electronic processor is configured to identify and isolate a plurality of conversations of interest from among the audio conversations based on predetermined characteristics indicative of a potential threat. The electronic processor is configured to determine that at least two of the plurality of conversations of interest are potentially linked to a single potential threat, compile them for playback or transcription, and generate an alert indicating the availability of the linked conversations of interest.

Abstract: Methods, apparatuses, system, devices, and computer program products for customized integrated entity analysis using an AI model are disclosed. In a particular embodiment, customized AI-powered analysis of an entity includes a controller creating a custom attribute for a user to analyze an entity and processing a first set of external data related to the created custom attribute and the entity including structured data and unstructured data retrieved from a first set of structured data sources and unstructured data sources. In this embodiment, the controller augments an AI model using the processed first set of external data and generates using the augmented AI model, one or more metrics for assessing the custom attribute in relation to the entity. The controller also presents to the user the generated one or more metrics for assessing the custom attribute in relation to the entity.

Abstract: The invention disclosed is an improved audio output module for use with an artificial voice generation device, having a housing separated into a sound system chamber, an interface chamber, and a power source chamber. The interface and power source chambers may be combined. The sound chamber is isolated from external air by the housing, the cover plate, and a separating wall, which separates it from other chambers of the module. Volumetric parameters based on speaker characteristics and design requirements can thus be implemented independent from the choice of interface type. The module is configurable to be mounted to an external structure or to a speech generating system. It may likewise be detachable from a quick release cradle and receive wireless audio signals from the speech generating system.

Abstract: Provided herein are system, apparatus, device, method, and/or computer program product embodiments, and/or combinations and sub-combinations thereof, for classifying a document using CNN and BiLSTM.

Abstract: This disclosure provides methods, devices, and systems for signal processing. The present implementations relate more specifically to a spatio-temporal beamformer. In some aspects, a beamforming system may receive an audio signal via a plurality of microphones, the audio signal including a number (B) of frames for each of the plurality of microphones, each of the B frames for each of the plurality of microphones including a number (N) of time-domain samples. For a first microphone, the beamforming system may transform the B*N time-domain samples into B*N/2 first frequency-domain samples; transform the B*N/2 first frequency-domain samples into B*N/2 second frequency-domain samples; and determine a probability of speech associated with the B*N/2 second frequency-domain samples based on a neural network model. The beamformer system may determine a minimum variance distortionless response (MVDR) beamforming filter based at least in part on the probability of speech for the first microphone.

Abstract: Systems and methods are provided for sound signal processing using neuromorphic analog signal processors. A hardware apparatus includes a digital switch coupled to a plurality of analog neuromorphic cores. The digital switch is configured to obtain one or more sound streams from one or more sound sources, transmit data based on the one or more sound streams to the plurality of analog neuromorphic cores, receive output from the plurality of analog neuromorphic cores, and output one or more modified sound streams based on the output received from the plurality of analog neuromorphic cores. Each analog neuromorphic core includes a respective analog network of analog components and is configured to (i) receive respective input data from the digital switch, (ii) perform a respective voice-related function on the respective input data, and (iii) transmit respective output to the digital switch.

Abstract: According to various example embodiments, an electronic device includes a microphone configured to receive an audio signal including speech of a user, a processor, and a memory configured to store instructions executable by the processor and personal information of the user, in which the processor is configured to extract a plurality of speech recognition candidates by analyzing a feature of the speech of the user, extract a keyword based on the plurality of speech recognition candidates, search for replacement data, based on the keyword and the personal information, and generate a recognition result corresponding to the speech of the user, based on the replacement data.

Abstract: A method of training, by a terminal, a mathematics-related artificial intelligence (AI) model, includes collecting mathematical words for training the mathematics-related AI model, converting the mathematical words into a reading sentence by using a mathematical reading technology, generating learning data for training the mathematics-related AI model based on the reading sentence, and training the mathematics-related AI model by using the learning data. The mathematical words may include a natural language part and a mathematical part.

Abstract: The clothing coordination system may comprise a tag and a tag manager. The clothing coordination system may be adapted to audibly identify a tagged item for a user. The tag may be coupled to the tagged item. The tag manager may be adapted to digitize a user's voice as the user verbally describes the tagged item. The tag manager may create a digital representation of the user's voice and may wirelessly transmit the digital representation to the tag where the digital representation May be stored within the tag. The tag manager may further be adapted to wirelessly receive the digital representation from the tag and to reproduce the user's voice as audio played through a sound transducer on the tag manager.

Abstract: Aspects disclosed herein generally related to a method and system for efficient blind source separation using a topological approach. The method and system comprise locating and separating the audio streams by constructing and simplifying contour tree in a built time-frequency smooth weighted histogram in the subsystems included. Thus, in one example, the audio streams can be separated and reproduced in a faster, more reliability, higher quality and more robust way.

Abstract: A wearable electronic device (WED) worn on a head of a person determines a location of the person, retrieves a room impulse response (RIR) for the location when the RIR is stored in memory, generates the RIR for the location when the RIR is not stored in the memory, and plays sound processed with the RIR to the person.

Abstract: One aspect of the present disclosure relates to a method of sentiment analysis based on ambiguity analysis, which includes analyzing information with the sentiment analysis models and the ambiguity analysis models. Another aspect of the present disclosure relates to a method of training the sentiment analysis models and ambiguity analysis models, which includes acquiring information, constructing lexicons, conducting sentiment analysis and ambiguity analysis with said lexicons, acquiring corpus, and training models, etc. Meanwhile, another aspect of the present disclosure relates to a system of sentiment analysis, which includes input, and output modules, acquisition modules, processing modules and database.

Abstract: The present disclosure relates to an information system, a method of providing information, and a respective electronic device. The electronic device is operable to provide an information to a user being indicative of a user device. The electronic device comprises: a processor to process a user request, a user interface to communicate with the user, and an electronic storage connected to the processor. The processor is operable: to extract at least one keyword from the user request, to select at least one information content from a content database on the basis of the at least one extracted keyword, and to provide the at least one selected information content to the user via the user interface.

Abstract: The disclosure provides technology for enhancing the ability of a computing device to detect when a user has discontinued reading a text source. An example method includes receiving audio data comprising a spoken word associated with a text source, comparing the audio data with data of the text source, determining, based on the comparing, whether a segment of the audio data corresponds to a location of the text source, and responsive to determining that the segment of the audio data does not correspond to a location of the text source, transmitting a signal indicating that a user has discontinued reading the text source, the signal causing to cease the comparing of the audio data with the data of the text source.

Abstract: Innovations in phase quantization during speech encoding and phase reconstruction during speech decoding are described. For example, to encode a set of phase values, a speech encoder omits higher-frequency phase values and/or represents at least some of the phase values as a weighted sum of basis functions. Or, as another example, to decode a set of phase values, a speech decoder reconstructs at least some of the phase values using a weighted sum of basis functions and/or reconstructs lower-frequency phase values then uses at least some of the lower-frequency phase values to synthesize higher-frequency phase values. In many cases, the innovations improve the performance of a speech codec in low bitrate scenarios, even when encoded data is delivered over a network that suffers from insufficient bandwidth or transmission quality problems.

Abstract: Enabling machine-to-machine communication for digital assistants can include initiating a call with a called device, generating audio stream data having a first instance of audio and a first spoken command, which can be provided to the called device. A second instance of audio stream data can be received from the called device, can include a second spoken command, and can be analyzed to determine if it includes assistant signature data. If the second instance of audio stream data includes the assistant signature data, the devices can switch to machine-to-machine communications.