Feature Gate Paralinguistic Estimation for Noise Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing paralinguistic information estimation models struggle to accurately learn utterances where characteristics appear only in one or some features, leading to reduced accuracy due to noise from features where characteristics are not present.
Innovation Solution
A paralinguistic information estimation apparatus that includes feature gates to determine the relevance of each feature for estimation, allowing the model to selectively use features based on the prominence of characteristics, thereby improving the learning and estimation process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a paralinguistic information estimation model uses a plurality of independent features for estimation, then the accuracy of paralinguistic information estimation is improved, but the model learns noise from features where characteristics are not present, reducing accuracy
Solution Approach 1:
The patent applies dynamics by making the feature selection adaptive rather than static. The determination unit dynamically identifies which features actually contain paralinguistic information characteristics for each specific utterance, allowing the model to adapt its feature usage based on the input data rather than always using all available features.
Solution Approach 2:
The patent applies local quality by treating different features differently based on their actual content. Instead of uniformly using all features or none, the system selectively applies features that contain relevant characteristics for each specific case, making the feature usage heterogeneous and tailored to the local characteristics of each utterance.
2Device complexity
If the model assumes all features indicate the same characteristics of paralinguistic information, then the model structure is simplified, but utterances with characteristics appearing only in one or some features cannot be correctly learned
Solution Approach 1:
The patent transforms the static assumption that all features indicate the same characteristics into a dynamic determination process. The determination unit dynamically assesses which features actually reflect paralinguistic information for each utterance, allowing the model to handle diverse feature characteristics without requiring complex manual configuration.
Solution Approach 2:
The patent applies self-service by enabling the model to automatically determine which features are relevant through the determination unit. The system self-adjusts its feature selection based on the actual characteristics present in the input data, eliminating the need for external manual feature selection or complex pre-processing.
Data Source
AI summary
To increase the accuracy of paralinguistic information estimation. A paralinguistic information estimation model storage unit 20 stores a paralinguistic information estimation model outputting, with a plurality of independent features as inputs, paralinguistic information estimation results. A feature extraction unit 11 extracts the features from an input utterance. A paralinguistic information estimation unit 20 estimates paralinguistic information of the input utterance from the features extracted from the input utterance, by using the paralinguistic information estimation model. The paralinguistic information estimation model includes, for each of the features, a feature sub-model outputting information to be used for estimation of paralinguistic information, based only on the feature, for each of the features, a feature weight calculation unit calculating a feature weight, based on an output result of the feature sub-model, for each of the features, a feature gate weighting the output result from the feature sub-model with the feature weight and outputting a result, and a result integration sub-model estimating the paralinguistic information, based on output results from all the feature gates.