Automatic Call Screening With Audio Embeddings for Spoofed Numbers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Spam calls, particularly robocalls, are a significant problem due to the use of fake or spoofed numbers, which undermines the efficacy of number-based anti-spam tools, leading to user inconvenience and increased interaction with unwanted calls.
Innovation Solution
A computer-implemented method using a trained machine-learning model generates embeddings from audio recordings and metadata of calls, clusters similar calls, and automatically determines and mitigates spam calls on client devices without user intervention, utilizing local storage and processing to compare call embeddings with spam models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If number-based anti-spam tools are used, then call filtering is simplified, but efficacy deteriorates due to fake or spoofed numbers
Solution Approach 1:
The patent replaces number-based filtering (mechanical system) with voiceprint-based biometric authentication. Instead of comparing phone numbers, the system extracts and compares unique vocal characteristics from call audio, making the filtering mechanism immune to spoofed numbers while maintaining operational simplicity.
Solution Approach 2:
The patent changes the identification parameter from phone number (metadata) to voiceprint characteristics (audio features). By analyzing spectral features, pitch, timbre, and other vocal parameters, the system creates a reliable identifier that cannot be faked, directly resolving the efficacy problem while keeping the interface simple.
2Measurement precision
If all calls are manually reviewed, then spam detection accuracy is maximized, but user time and productivity deteriorate
Solution Approach 1:
The patent implements self-service spam detection where the system automatically performs voiceprint extraction, comparison, and call classification without user intervention. The automated system maintains high accuracy by using sophisticated audio analysis while eliminating the need for manual review, thus saving user time.
Solution Approach 2:
The system incorporates feedback mechanisms where user responses to automatic classifications are used to refine and update voiceprint models continuously. This feedback loop maintains high detection accuracy over time while keeping the process automated, preventing the need for manual review.
3Reliability
If voice-based authentication is implemented, then spam call identification improves, but device complexity increases
Solution Approach 1:
The patent segments the voice authentication process into distinct modules: audio recording, pre-processing, feature extraction, voiceprint generation, comparison, and decision-making. Each module handles a specific task, making the overall complex system manageable and maintainable while achieving high authentication reliability.
Solution Approach 2:
The patent introduces voiceprint templates as an intermediary representation between raw audio and authentication decisions. These pre-generated templates serve as a bridge, allowing complex audio analysis to be performed once during setup and then used for rapid, simpler comparisons during actual call authentication.
4Reliability
If user privacy is fully protected, then data security improves, but information availability for spam detection deteriorates
Solution Approach 1:
The patent extracts only the essential voiceprint features from call audio while discarding all other personal information. By taking out only the necessary biometric data (spectral characteristics, pitch patterns) and leaving behind identifying information like phone numbers or personal details, the system maintains data security while preserving enough information for effective spam detection.
Data Source
AI summary
Implementations described herein relate to methods, systems, and computer-readable media to automatically answer a call. In some implementations, a method includes receiving a call from a caller device at a client device. The method further includes determining, based on an identifier associated with the call, whether the call matches auto answer criteria, and yin response to determining that the call matches the auto answer criteria, answering the call without user input and without alerting a user of the client device. The method further includes generating a call embedding for the call based on received audio of the call, comparing the call embedding with spam embeddings to determine whether the call is a spam call, and in response to determining that the call is a spam call, terminating the call.