Back to results
US 2026/0192178 A1Published application
Action control system and information processing system
AssigneeSoftBank Group Corp.
InventorMasayoshi SON
Life of this patent
Estimating the term…
Abstract
An avatar is caused to perform an action appropriate for an action of a user. In an action control system, in a case in which an action of the avatar includes dreaming and the action determination unit determines dreaming, as the action of the avatar, the action determination unit creates an original event obtained by combining multiple pieces of event data among pieces of data in history data.
Claims20
A multi-sensor detection apparatus comprising:
- a sensor assembly including:
- an image sensor configured to capture visible-spectrum image data of a detection zone at a first frame rate,
- a depth sensor configured to emit structured infrared light patterns and capture infrared reflections to generate three-dimensional depth data of the detection zone, and
- a motion sensor configured to detect movement within the detection zone;
- an image processing pipeline including:
- a feature extraction module configured to:
- apply a convolutional neural network to the image data to extract feature vectors representing objects within the detection zone, and
- identify a plurality of entities based on the extracted feature vectors,
- a tracking module configured to:
- correlate the three-dimensional depth data with the extracted feature vectors, and
- generate position data for each of the identified entities, and
- a behavior analysis module configured to:
- analyze temporal sequences of the position data to identify behavioral patterns of the identified entities;
- a storage device configured to store the behavioral patterns;
- and
- processing circuitry configured to:
- receive a query regarding a specific entity among the identified entities,
- retrieve behavioral patterns associated with the specific entity from the storage device,
- generate advice content based on the retrieved behavioral patterns using a sentence generation model, and
- output the advice content via an avatar rendered on a display device.
17 more elements
A multi-sensor detection apparatus comprising:
- a sensor assembly including:
- an image sensor configured to capture visible-spectrum image data of a competition space at a capture frame rate,
- a depth sensor configured to emit structured infrared patterns and capture infrared reflections to generate three-dimensional position data of players within the competition space, and
- an acceleration sensor configured to detect motion data;
- an image processing pipeline including:
- a feature extraction module configured to apply a neural network to the image data to extract feature vectors representing the players,
- a tracking module configured to generate position data for each player by correlating the three-dimensional position data with the feature vectors, and
- a behavior analysis module configured to analyze temporal sequences of the position data to identify movement habits, mistake frequencies, and reaction speeds of the players;
- a storage device configured to store the identified movement habits, mistake frequencies, and reaction speeds;
- and
- processing circuitry configured to:
- identify, from the feature vectors, a player belonging to an opponent team, retrieve behavioral data associated with the identified player from the storage device,
- generate competition advice content based on the retrieved behavioral data using a sentence generation model,
- render an avatar on a display device, the avatar having an appearance resembling the identified player, and
- output the competition advice content via the rendered avatar.
12 more elements
A method for generating competition advice based on multi-sensor detection, the method comprising:
- capturing, by an image sensor, visible-spectrum image data of a competition space;
- emitting, by a depth sensor, structured infrared light patterns and capturing infrared reflections to generate three-dimensional depth data of the competition space;
- applying, by a feature extraction module, a convolutional neural network to the image data to extract feature vectors representing players within the competition space;
- correlating, by a tracking module, the three-dimensional depth data with the extracted feature vectors to generate position data for each player;
- analyzing, by a behavior analysis module, temporal sequences of the position data to identify behavioral patterns including movement habits and mistake frequencies;
- storing the behavioral patterns in a storage device;
- receiving a query regarding a specific player among the players;
- retrieving behavioral patterns associated with the specific player from the storage device;
- generating, by applying a sentence generation model to the retrieved behavioral patterns, competition advice content;
- and
- outputting the competition advice content via an avatar rendered on a display device.
8 more elements
Dependencies are inferred from the claim text.