Vocoder Type Identification via Machine Learning Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In voice communication, the lack of prior information about the vocoder used for encoding a voice signal leads to inefficient decoding, as all available decoders must be used to restore the signal, significantly reducing decoding efficiency.
Innovation Solution
A method and apparatus utilizing a machine learning model to identify the type of vocoder used for encoding by acquiring and analyzing bitstreams and waveforms, employing multiple machine learning models to determine the vocoder type based on probability and intelligibility values, thereby enabling efficient decoding without prior information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all available decoders are used to restore the voice signal without prior information about the vocoder type, then the voice signal can be decoded, but the decoding efficiency is significantly reduced
Solution Approach 1:
The system performs preliminary action by training multiple machine learning models in advance to identify vocoder types. Before actual decoding occurs, the models are ready to quickly determine which vocoder was used to encode the bitstream, allowing the receiver to select the appropriate decoder without having to try all available decoders, thus resolving the contradiction between reliability and productivity
Solution Approach 2:
Machine learning models serve as an intermediary between the received bitstream and the decoder selection process. The models analyze the bitstream and provide probabilistic information about the vocoder type, enabling intelligent intermediary decision-making that avoids the inefficiency of trying all decoders while ensuring accurate decoding through model-based prediction
2Productivity
If multiple machine learning models are used to identify the vocoder type with high accuracy, then the decoding efficiency is improved, but the system complexity increases
Solution Approach 1:
The system segments the complex vocoder identification task into multiple specialized machine learning models, each trained to identify specific vocoder types or characteristics. This segmentation allows each model to be relatively simple and focused, while collectively they provide comprehensive and accurate vocoder type identification, resolving the contradiction between productivity improvement and system complexity
Solution Approach 2:
The system uses parameter changes by training models with different input parameters and features to identify various vocoder types. Each model is optimized for specific parameters or characteristics of certain vocoders, allowing the system to achieve high decoding efficiency through parameter-based model selection without requiring a single overly complex model
Data Source
AI summary
In accordance with an aspect of the present disclosure, there is provided a method for identifying a type of a vocoder. The method comprises acquiring identification target bitstreams encoded with a voice signal, acquiring, for each of a plurality of vocoders, a probability that each of the plurality of vocoders is related to the identification target bitstreams from the identification target bitstreams, acquiring waveforms for each decoder of each of the plurality of vocoders by inputting the identification target bitstreams to the each decoder of each of the plurality of vocoders, acquiring intelligibility values for each of the waveforms obtained for the each decoder of each of the plurality of vocoders from the waveforms, and determining the type of the vocoder related to the voice signal from the probability and the intelligibility values for each waveform.