Audio Encoder Selection Using Dual Classifier Decision Logic

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing audio encoding technologies face challenges in accurately classifying audio frames as speech or non-speech, leading to misclassification and subsequent artifacts or poor signal quality due to the limitations of Gaussian mixture models and open-loop classifiers.

Innovation Solution

A system that employs multiple classifiers, including a model-based classifier and an open-loop classifier, to generate decision data for selecting between speech and non-speech encoders, using adjustment parameters to bias the selection and improve accuracy by considering factors like sample rate, coding gain, and voicing values.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a single classifier (GMM or open-loop) is used to classify audio frames, then the classification process is simple and fast, but the classification accuracy is poor leading to misclassification

Engineering Contradiction:
Improveclassification accuracyVSAvoidclassifier structure
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent combines multiple classifiers (GMM-based classifier and open-loop classifier) into a unified classification system. The GMM-based classifier analyzes speech characteristics like pitch and spectral shape, while the open-loop classifier predicts encoder suitability. By merging their outputs through a decision logic that considers both classifiers' results, the system achieves higher classification accuracy than either classifier could achieve alone, resolving the contradiction between simplicity and accuracy.

Inventive Principle:
Principle #5Merging (Combining)

2Reliability

If misclassification occurs in audio frame classification, then the encoding process is fast, but artifacts and poor signal quality are produced

Engineering Contradiction:
Improvesignal qualityVSAvoidclassification system
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent implements a feedback mechanism where the classification system continuously monitors audio frame characteristics and adjusts encoder selection based on classification confidence. When the combined classifier output indicates high confidence in speech or non-speech classification, the system proceeds with encoding. This feedback loop ensures that only reliable classifications lead to encoding decisions, preventing artifacts and poor signal quality while maintaining appropriate system complexity.

Inventive Principle:
Principle #23Feedback

3Measurement precision

If multiple classifiers are used to improve classification accuracy, then misclassification is reduced, but the processing complexity and computational load increase

Engineering Contradiction:
Improveframe classification accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies partial action by using the GMM-based classifier as the primary classification mechanism for most audio frames, and selectively engaging the open-loop classifier based on specific conditions or confidence thresholds. This approach achieves improved classification accuracy where needed without applying the full computational power of both classifiers to every frame, thereby reducing overall processing time while maintaining high accuracy for critical classification decisions.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3281198B1Encoder selection
Publication Date: 2019.11.06 QUALCOMM INC
  • EP3281198B1 patent drawingFigure 1
  • EP3281198B1 patent drawingFigure 2
  • EP3281198B1 patent drawingFigure 3

AI summary

A device includes a first classifier and a second classifier coupled to the first classifier. The first classifier is configured to output first decision data that indicates a classification of an audio frame as a speech frame or a non-speech frame, the first decision data determined based on first probability data associated with a first likelihood of the audio frame being the speech frame and based on second probability data associated with a second likelihood of the audio frame being the non-speech frame. The second classifier is configured to output second decision data based on the first probability data, the second probability data, and the first decision data, the second decision data includes an indication of a selection of a particular encoder of multiple encoders available to encode the audio frame.