Home-Specific Sound Event Detection Using Generative Probabilistic Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current security systems are limited in distinguishing between human and pet sounds, and cannot accurately determine if a detected sound is specific to a particular home, leading to potential false alarms and lack of personalized security event detection.
Innovation Solution
A smart home environment that uses a sensor network, including audio sensors, to detect sound events, classify them into human or pet-generated sounds, and transmit notifications using a generative probabilistic model that learns home-specific sound patterns, considering room size and reverberation to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a sensor compares detected sound with pre-stored sound to determine security events, then the system can detect sound events, but it cannot distinguish between human and pet sounds and produces false alarms
Solution Approach 1:
The sound detection system is segmented into multiple independent analysis components: a sound event detector that identifies potential security events, a human sound detector that analyzes fundamental frequency and harmonics to determine human presence, and a pet sound detector that analyzes pitch and temporal patterns to identify pet sounds. This segmentation allows the system to process sound data through multiple specialized detectors, each optimized for specific sound source identification, thereby resolving the inability to distinguish between different sound sources while maintaining reliable security event detection
Solution Approach 2:
The processor acts as an intermediary that receives sound data from the sensor and routes it through multiple detection algorithms. The processor analyzes the sound data using both human sound detection algorithms (examining fundamental frequency, harmonics, and spectral characteristics) and pet sound detection algorithms (examining pitch, temporal patterns, and frequency ranges). This intermediary processing layer enables the system to determine whether a detected sound event is human-generated, pet-generated, or another source, thereby eliminating false alarms while preserving accurate security event detection
2Adaptability or versatility
If the system uses a single pre-stored sound template for all homes, then the system is simple to implement, but it cannot learn home-specific sound patterns
Solution Approach 1:
The system performs preliminary sound data collection during a learning period after installation, gathering sound samples from the specific home environment. During this preliminary phase, the processor analyzes sounds from various sources (humans, pets, appliances, external environments) and builds home-specific sound profiles. This preliminary action enables the system to adapt to each home's unique acoustic characteristics, room size, reverberation properties, and typical sound sources before normal operation begins, thereby achieving home-specific pattern recognition without requiring complex real-time adaptation mechanisms
Solution Approach 2:
The sound detection system dynamically adjusts its parameters and thresholds based on home-specific learning. The processor continuously refines the human and pet sound detection algorithms by comparing detected sounds against the learned home-specific profiles. This dynamic adaptation allows the system to optimize its sensitivity and specificity for each particular home environment, improving adaptability while managing complexity through automated learning rather than manual configuration
Data Source
AI summary
Systems and methods of a security system are provided, including detecting, by a sensor, a sound event, and selecting, by a processor coupled to the sensor, at least a portion of sound data captured by the sensor that corresponds to at least one sound feature of the detected sound event. The systems and methods include classifying the at least one sound feature into one or more sound categories, and determining, by a processor, based upon a database of home-specific sound data, whether the at least one sound feature is a human-generated sound. A notification can be transmitted to a computing device according to the sound event.