Distributed Voice Recognition Across Multiple Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face challenges in maintaining consistent performance of voice and sound recognition across multiple devices, especially when disconnected from common services or cloud computing, due to variations in sensor sensitivity and signal processing characteristics.
Innovation Solution
A distributed system and method that enables multiple devices to collaborate by sharing samples, features, recognition scores, and risk scores to improve signal quality and synchronization, using a common algorithm model that can be trained and updated across devices, optimizing tasks based on processing power, bandwidth, and sensor quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If voice recognition is performed disconnected from cloud services across multiple devices, then authentication can be performed locally without network dependency, but recognition accuracy deteriorates due to variations in sensor characteristics and lack of centralized model updates
Solution Approach 1:
The patent combines recognition scores from multiple devices to make authentication decisions. Each device independently performs voice recognition using local models, and the system merges the individual recognition scores into a collaborative decision, thereby maintaining authentication reliability without cloud dependency while improving accuracy through multiple observations.
Solution Approach 2:
The patent distributes identical voice recognition models to multiple devices, creating copies of the authentication system across the network. Each device maintains a local copy of the model and can perform independent recognition, ensuring consistency and accuracy across devices while operating disconnected from centralized services.
2Stability of the object's composition
If multiple devices use identical voice recognition algorithms, then authentication consistency is maintained across devices, but performance deteriorates due to variations in microphone sensitivity and signal processing characteristics
Solution Approach 1:
The patent applies local quality adjustments by calibrating each device's recognition output according to its specific microphone characteristics. Each device's recognition score is adjusted based on its individual sensor properties, allowing consistent authentication decisions across devices while compensating for hardware variations to maintain recognition performance.
Solution Approach 2:
The patent changes the parameters of the recognition system by introducing device-specific calibration factors and score adjustments. Rather than using identical algorithms unchanged, the system modifies operational parameters for each device based on its characteristics, achieving both consistency in authentication decisions and performance adaptation to hardware variations.
3Measurement precision
If devices share recognition data collaboratively, then recognition accuracy improves through multiple observations, but system complexity increases due to inter-device communication and score fusion requirements
Solution Approach 1:
The patent extracts only the essential recognition scores from each device for collaborative decision-making, rather than sharing complete audio data or complex model parameters. This extraction approach improves recognition accuracy through multiple observations while minimizing system complexity by transmitting only the necessary numerical results for fusion.
4Measurement precision
If voice recognition models are updated centrally, then all devices benefit from improved accuracy, but devices become dependent on network connectivity and centralized services
Solution Approach 1:
The patent performs preliminary actions by pre-distributing voice recognition models to all devices before they are needed for authentication. Devices receive and store model updates in advance during network availability, enabling them to operate independently when disconnected while still benefiting from centralized improvements when connected.
Solution Approach 2:
The patent introduces dynamics to the model update process by allowing devices to operate with local models during disconnection and automatically synchronize updates when reconnected. This dynamic approach balances recognition accuracy improvements from centralized updates with operational independence during network unavailability.
Data Source
AI summary
A distributed system and method to improve collaborative service across multiple sensors on various devices. According to one embodiment, multiple devices may be used to train and then utilize a common algorithm for purposes including but not limited to recognizing a source to perform some action, control, command, calculation, storage, retrieval, encryption, decryption, alerting, alarming, notifying or as in some embodiments, to authenticate. In one embodiment of the invention, devices with one or more sensors such as but not limited to microphones, acoustic arrays or audio sensors may contribute to one or more models by sending samples, features, recognition scores, and/or risk scores to each other to improve collaborative training, signal quality, recognition, synchronization, inter-device proximity location and/or fusion of recognition scores and/or risk scores.