Voice Call Classification via Biometric and Content Analysis

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current systems are ineffective in classifying calls from malicious entities, particularly those using number swapping technology, which can lead to legitimate subscribers being incorrectly blocked, disrupting communication between businesses and their customers.

Innovation Solution

A method for voice call classification that intercepts and analyzes voice call data, transforming it into a predefined format to identify attributes such as call category, emotional components, and presence of robotized speech, using machine learning algorithms like naive Bayesian classifiers and SVM to classify calls as regular or unwanted, and notify the user accordingly.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Object-affected harmful factors

If conventional systems block numbers based on database matching, then spam calls are filtered, but legitimate calls with swapped numbers are incorrectly blocked

Engineering Contradiction:

Improvespam call filteringVSAvoidfalse blocking of legitimate calls

Core Design Contradiction:

Object-affected harmful factorsVSReliability

Solution Approach 1:

The system segments the call classification task into multiple analysis dimensions: voice biometric analysis to verify caller identity, content analysis to detect marketing/fraud patterns, and metadata analysis to identify number swapping. This multi-dimensional segmentation allows the system to distinguish between malicious calls using swapped numbers and legitimate calls, reducing false positives while maintaining spam filtering effectiveness

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system performs preliminary voice biometric analysis and content analysis during the call setup phase before the call is fully established. By analyzing voiceprints, detecting robotized speech patterns, and identifying marketing content in advance, the system can proactively block malicious calls while allowing legitimate calls to proceed, preventing false blocking of legitimate business communications

Inventive Principle:

Principle #10Preliminary action

2Object-generated harmful factors

If number swapping technology is used by hackers, then call anonymity is achieved, but legitimate subscribers are discredited

Engineering Contradiction:

Improvecall anonymity for hackersVSAvoiddiscrediting of legitimate subscribers

Core Design Contradiction:

Object-generated harmful factorsVSLoss of information

Solution Approach 1:

The system introduces voice biometric analysis as an intermediary verification mechanism between the caller and the called party. By analyzing unique voice characteristics and comparing them against known patterns, the system can identify the true identity of the caller even when number swapping is used. This intermediary analysis prevents legitimate subscribers from being discredited by hackers using swapped numbers, while maintaining call anonymity for genuine privacy-protected legitimate calls

Inventive Principle:

Principle #24Intermediary (Mediator)

3Measurement precision

If voice call data is analyzed in real-time, then call classification accuracy is improved, but processing time increases

Engineering Contradiction:

Improvecall classification accuracyVSAvoidcall processing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary analysis of voice biometric features, content patterns, and metadata during the call setup phase before the full call is established. By preprocessing and analyzing key features in advance, the system can make rapid classification decisions without requiring extensive real-time processing during the actual call, thus maintaining high accuracy while minimizing processing time and avoiding call delays

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system implements a two-stage analysis approach: first performing a quick partial analysis of key features (voice biometrics, content keywords, metadata) to make an initial classification decision, and only if needed, performing more extensive analysis. This partial action approach achieves sufficient classification accuracy for most calls while minimizing processing time, avoiding excessive analysis that would delay legitimate calls

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS11380303B2System and method for call classification

Publication Date: 2022.07.05 AO KASPERSKY LAB

AI summary

A method for voice call analysis and classification includes intercepting a voice call session between an initiating device and a recipient device. Voice call data exchanged between the initiating device and the recipient device during the voice call session is transformed into a predefined data format. The transformed voice call data is analyzed to determine one or more attributes of the intercepted voice call. One or more features associated with the intercepted voice call session are identified based on the determined one or more attributes. The intercepted voice call is classified using the identified one or more features.