AI & Machine Learning

Sony received 16 patents covering a wide range of AI-powered gaming technologies. The collection includes systems that extract structured metadata from legacy games by analyzing controller inputs, video frames, and audio, enabling modern console features to work without developer APIs or source code access. Several patents address player guidance and personalization, including a recommendation engine that suggests optimal in-game items based on player behavior and community data, a server-side system that aggregates player data to generate crowd-sourced gameplay help showing which mechanics work best for specific challenges, and an AI that analyzes NPC interactions across gaming sessions to customize dialogue based on individual play patterns. Others focus on content creation and accessibility, such as a character creation system combining motion capture from simple video with natural language descriptions, an avatar generator that transforms selfies into game-ready characters matching each game's art style, a voice synthesis system that automatically generates audio descriptions in gaps between game dialogue, and a dynamic music generator that creates personalized soundtracks in real-time based on player interactions and gameplay state. The company also patented social and streaming features, including an AI gaming streamer that adjusts its gameplay based on viewer reactions to maximize audience engagement rather than simply winning, a message filtering system that rewrites in-game communications to fit each recipient's current gameplay context while preserving intent, a gesture recognition system that translates sign language into context-aware chat messages, and a voice-controlled system that detects multiple players' voices simultaneously through spatial audio processing. Additional patents cover biometric integration that captures player stress levels to enhance spectator viewing and create emotionally-responsive NPCs, matchmaking that predicts gaming session duration based on player history to help find compatible teammates, a video analysis tool that predicts audience engagement for game trailers before release, and a context-aware gesture and voice input system that uses game state data to create dynamic time windows where input is appropriate, preventing accidental commands during natural movements.

EA received 4 patents focused on AI-driven character animation and player personalization. Two address animation generation, with one creating real-time poses procedurally without storing motion capture data by learning the grammar of human movement during training, and another automatically generating expressive facial animations from speech audio by recognizing emotional tone to eliminate manual animation work. A third uses neural network clustering to map player behavior types and automatically deliver personalized in-game guidance, combining self-organizing maps with reinforcement learning to monitor behavioral migration across player clusters and adjust strategies in real-time. The fourth applies geometric analysis of game environments to generate context-aware assistance that helps players navigate obstacles and make tactical decisions based on spatial relationships between objects.

Nintendo received 1 patent for creating viral wave effects in games where NPCs progressively imitate player actions in expanding circles. Instead of all NPCs reacting simultaneously, the system creates a wave-like propagation effect where characters in expanding concentric zones start imitating at different times based on distance plus randomized waiting periods, with optional background music synchronization to maintain rhythm cohesion.

ByteDance received 1 patent for an AI-powered matchmaking system designed specifically for MOBA games. The system clusters players by skill level across multiple pools to reduce wait times while improving match quality, training the clustering model with dual constraints that ensure game level differences between teams across different class clusters while minimizing quantitative differences between pools.

Ovomind received 1 patent for predicting player emotional responses to video game sequences. The system combines multi-modal game content analysis, including audio via convolutional neural networks, video via colorimetric histograms, and graphics via classifiers, with timestamped biosignal processing to create predictive emotional models that work prospectively rather than reactively.

Colopl received 1 patent for an AI-powered character interaction system that generates dynamic, context-aware NPC responses. Instead of selecting from pre-programmed dialogue trees, the system uses machine learning to create character responses that weren't explicitly pre-authored, allowing NPCs to exhibit behaviors beyond their original programming.

Cygames received 1 patent for automated quality assurance in complex multiplayer games. The system analyzes game logs to identify specific card combinations and game state patterns causing bugs, automatically determining causality by statistically comparing the frequency of specific action-state combinations in failed versus successful sessions to find problems that only occur when certain elements appear together.

Gna received 1 patent for a machine learning system that clusters game players by behavior patterns to enable targeted marketing. The system moves beyond demographic targeting to behavioral clustering based on actual gameplay patterns and in-game actions, allowing precision targeting of game advertisements to players most likely to engage with specific game types.

Microsoft received 1 patent for AI-powered game development where generative models create narrative content from natural language prompts. The system functions as a co-pilot for designers during development, translating natural language to game content, plus generates content in real-time during gameplay that adapts based on implicit player feedback metrics, creating an evolving experience where players become co-creators.

GDM Holding received 1 patent for multi-agent reinforcement learning that trains AI agents to master complex tasks by learning from each other. The matchmaking policy framework intelligently pairs agents during training, allowing different learner policies to use different strategies and hyperparameters while learning from each other's behaviors, creating a collectively improving population rather than training in isolation.

Hangzhou Electronic Soul Network Technology received 1 patent for automatically detecting toxic player behavior in PVP games. The system analyzes gameplay, performance, and social data to identify trolling, griefing, and AFKing without relying on player reports, shifting from reactive reporting systems to proactive automated detection using multi-dimensional analysis during a single match.

Apple received 1 patent for an agent-agnostic motion planner that generates movement paths for virtual characters across different XR and VR applications. The system performs runtime analysis of motion controller characteristics to generate controller-compatible motion plans, eliminating the need for agent-specific motion planners by adapting to each character's unique control scheme and movement capabilities using rapidly-exploring random trees and machine-learned root motion models.

Dell received 1 patent for a gaming controller with adjustable center of gravity that dynamically shifts weight based on in-game events. Unlike conventional haptic feedback relying on vibration motors, the system provides directional weight shifting as a feedback mechanism, allowing developers to communicate in-game physics and events through the controller's actual balance and center of gravity changes.