Automatic Call Categorization with Audio-Based Spam Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Spam calls, particularly robocalls, are a significant problem due to the use of fake or spoofed numbers, making number-based anti-spam tools ineffective, and users receive a high volume of unwanted calls without effective automated filtering solutions.
Innovation Solution
A computer-implemented method using a trained machine-learning model generates embeddings from audio recordings and text transcripts of calls, groups them into clusters, and identifies spam calls on client devices without user intervention, allowing automatic answering or termination based on spam embeddings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If number-based anti-spam tools are used, then spam call filtering is simplified, but effectiveness deteriorates due to fake or spoofed numbers
Solution Approach 1:
The patent replaces number-based filtering (mechanical system) with audio-based fingerprinting and machine learning analysis. Instead of relying on caller ID numbers, the system records call audio, extracts acoustic features, and compares them against known spam patterns using trained models, thereby achieving effective spam detection despite number spoofing.
Solution Approach 2:
The patent changes the detection parameter from static caller ID numbers to dynamic audio characteristics. By analyzing acoustic features such as tone, pitch, speech patterns, and background noise in call recordings, the system creates a more reliable identifier that cannot be easily faked, resolving the effectiveness issue while maintaining reasonable complexity.
2Measurement precision
If manual spam call identification is used, then accuracy is maintained, but user time and productivity deteriorate due to high call volumes
Solution Approach 1:
The patent performs preliminary spam detection by analyzing call audio in real-time or near-real-time before the user needs to manually evaluate each call. The machine learning model pre-processes and categorizes calls, flagging potential spam automatically, which maintains high accuracy while eliminating the need for users to manually review every incoming call.
Solution Approach 2:
The system enables self-service spam filtering where the automated machine learning model independently identifies and flags spam calls without requiring user intervention. The model learns from training data and autonomously applies detection rules, preserving accuracy while freeing user time for legitimate calls only.
3Productivity
If automated spam filtering is implemented, then user time is saved, but system complexity increases due to machine learning model requirements
Solution Approach 1:
The patent segments the automated filtering system into modular components: audio recording module, feature extraction module, machine learning classification module, and result output module. This segmentation allows the complex system to be implemented incrementally and maintained separately, reducing the perceived complexity while maintaining high productivity through automation.
Solution Approach 2:
The machine learning model serves multiple functions: it classifies spam calls, identifies call patterns, and can be trained on diverse datasets to adapt to different spam types. This multi-functionality consolidates what could be multiple separate systems into a single versatile platform, improving productivity without proportionally increasing complexity.
4Measurement precision
If call audio analysis is performed, then spam detection accuracy is improved, but data processing requirements and energy consumption increase
Solution Approach 1:
The patent extracts only the most discriminative acoustic features from call audio (such as spectral characteristics, pitch contours, and speech rate) rather than processing the entire audio signal in detail. This selective extraction maintains high detection accuracy while significantly reducing computational load and energy consumption compared to full audio analysis.
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.