Ternary Audio Fingerprints for Low-Latency Network Message Matching

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current audio identification methods, particularly in call progress analysis, face challenges with high false positive rates and computational intensity, making them inefficient and inaccurate for real-time identification of network messages during outbound calls.

Innovation Solution

The development of a ternary fingerprint bitmap system, which creates a more efficient comparison process by combining multiple binary fingerprint bitmaps and masking out sensitive bits, allowing for faster and more accurate identification of audio patterns.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional binary fingerprint bitmap methods are used for audio identification, then the system can identify audio patterns, but the computational intensity is high and false positive rates are high

Engineering Contradiction:
Improveidentification accuracyVSAvoidcomputational intensity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent segments the binary fingerprint bitmap into multiple components and processes them separately. Specifically, it divides the bitmap into frequency sub-bands and temporal segments, allowing parallel processing and reducing the computational burden on any single processing unit while maintaining overall identification accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies different processing strategies to different regions of the fingerprint bitmap based on their local characteristics. Certain bits that are more prone to false positives are identified and treated differently (e.g., through weighting or selective validation), while stable bits are processed more efficiently. This local differentiation reduces overall computational intensity by avoiding uniform heavy processing across all bits.

Inventive Principle:
Principle #3Local quality

2Measurement precision

If traditional audio identification methods process entire audio segments, then identification can be performed, but the processing time is long and latency is high

Engineering Contradiction:
Improveidentification accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary processing on the audio signal before full identification is required. It extracts early features and creates preliminary fingerprints that can be used for quick preliminary identification or filtering. This allows the system to identify obvious matches quickly without waiting for complete audio segment processing, thereby reducing latency while maintaining accuracy for non-obvious cases.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements a staged processing approach where identification can proceed with partial audio segments when sufficient information is available. Instead of requiring complete audio segments for all identifications, the system can make accurate identifications with partial data in many cases, reducing average processing time while maintaining precision through confidence scoring and validation mechanisms.

Inventive Principle:
Principle #16Partial or excessive action

3Reliability

If AMD algorithms use heuristics and statistical models to classify utterances, then accuracy can be improved, but false positives increase and response time decreases

Engineering Contradiction:
Improveclassification accuracyVSAvoidfalse positives
Core Design Contradiction:
ReliabilityVSObject-generated harmful factors

Solution Approach 1:

The patent implements feedback mechanisms where the results of fingerprint matching are used to adjust and refine the AMD classification decisions. The system cross-validates AMD outputs against fingerprint match results, and uses this feedback to reduce false positives. When fingerprint evidence contradicts AMD classification, the system can override or flag the classification for review, thereby improving reliability while maintaining response time through automated feedback loops.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP3076391B1Creating a ternary bitmap of a dataset
Publication Date: 2018.07.04 INTERACTIVE INTELLIGENCE INC
  • EP3076391B1 patent drawingFigure 1A~1C
  • EP3076391B1 patent drawingFigure 2
  • EP3076391B1 patent drawingFigure 3A

AI summary

Systems and methods for the matching of datasets, such as input audio segments, with known datasets in a database are disclosed. In an illustrative embodiment, the use of the presently disclosed systems and methods is described in conjunction with recognizing known network message recordings encountered during an outbound telephone call. The methodologies include creation of a ternary fingerprint bitmap to make the comparison process more efficient. Also disclosed are automated methodologies for creating the database of known datasets from a larger collection of datasets.