Distributed Emotional Recognition System for Resource-Constrained Devices

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing emotional recognition systems for telecommunication networks are rigid and inflexible, making them unsuitable for personalized applications and inefficient on devices with limited processing resources like mobile phones and PDAs.

Innovation Solution

A distributed emotional recognition system that allows for personalized configuration and operation on devices with scarce processing resources, enabling users to select scenarios and configure the system to optimize emotional recognition based on language, accuracy, and channel conditions, with local and remote processing capabilities working together to improve recognition accuracy and efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If emotional recognition is performed using substantial data processing resources, then recognition accuracy is improved, but device complexity and resource consumption increase

Engineering Contradiction:
Improverecognition accuracyVSAvoidprocessing resource requirements
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The emotional recognition system is segmented into multiple independent modules: speech signal acquisition module, feature extraction module (extracting prosodic, spectral, and temporal features), emotion classification module, and configuration module. This segmentation allows each module to be optimized independently and enables distributed processing across client-server architecture, reducing the processing burden on individual devices while maintaining overall recognition accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

A feature extraction intermediary layer is introduced between speech signal acquisition and emotion classification. This intermediary extracts and preprocesses relevant features (prosodic features like pitch and energy, spectral features like MFCCs, and temporal features) locally on the client device, then transmits only these extracted features to the server for final classification. This reduces the amount of data that needs to be processed and transmitted, enabling accurate recognition on resource-constrained devices.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Ease of operation

If distributed architecture is used to reduce local processing requirements, then ease of operation on mobile devices is improved, but system complexity increases

Engineering Contradiction:
Improveoperability on mobile devicesVSAvoidsystem architecture complexity
Core Design Contradiction:
Ease of operationVSDevice complexity

Solution Approach 1:

The server component is designed with multi-functionality, handling not only emotion classification but also model training, feature optimization, and configuration management. The client device is equipped with a configuration module that can adaptively adjust processing parameters based on available resources. This universal design allows the distributed system to function as an integrated whole while maintaining simplicity at the client level.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system implements feedback mechanisms where the server analyzes recognition results and performance metrics, then adjusts and retransmits optimized emotion recognition models to client devices. The configuration module on client devices provides feedback about local processing capabilities and resource availability, enabling dynamic adaptation of the system architecture to balance simplicity and performance based on actual operating conditions.

Inventive Principle:
Principle #23Feedback

3Adaptability or versatility

If system is made customizable for different scenarios, then adaptability is improved, but device complexity increases

Engineering Contradiction:
Improvepersonalization capabilityVSAvoidconfiguration complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The configuration module is designed to be dynamic rather than static, automatically adjusting recognition parameters, feature extraction depth, and processing thresholds based on detected scenario types and available resources. The system can dynamically switch between different emotion recognition models and configurations without requiring manual reconfiguration, enabling high adaptability while maintaining operational simplicity through automated scenario detection and configuration selection.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP2122610B1Customizable method and system for emotional recognition
Publication Date: 2018.12.26 TELECOM ITALIA SPA
  • EP2122610B1 patent drawingFigure 1
  • EP2122610B1 patent drawingFigure 2
  • EP2122610B1 patent drawingFigure 3

AI summary

An automated emotional recognition system (100; 100a; 100b) is provided. The automated emotional recognition system is adapted to determine emotional states of a speaker based on the analysis of a speech signal (SS). The emotional recognition system comprises at least one server function (110; 110a; 110b) and at least one client function (120; 120a; 120b) in communication with the at least one server function for receiving assistance in the determining the emotional states of the speaker. The at least one client function includes an emotional features calculator (215) adapted to receive the speech signal (SS) and to extract therefrom a set of speech features (SF) indicative of the emotional state of the speaker. The emotional state recognition system further includes at least one emotional state decider (235; 705) adapted to determine the emotional state of the speaker exploiting the set of speech features based on a decision model (MODj; CMOD). The server function includes at least a decision model trainer (237; 735, 740) adapted to update the selected decision model according to the speech signal. The decision model to be used by the emotional state decider for determining the emotional state of the speaker is selectable based on a context of use of the recognition system.