Multi-speaker Recognition in Cocktail Party Problem

TL;DR

This paper introduces a statistical decision framework for recognizing multiple speakers in cocktail party scenarios, utilizing Gaussian assumptions, voiceprint features, and a novel constellation mapping to improve identification accuracy.

Contribution

It presents a new approach combining Gaussian-based statistical decision theory with Euclidean distance voiceprint analysis for multi-speaker recognition.

Findings

Effective voiceprint feature extraction using Mel-Frequency Cepstral Coefficients

Mapping of speaker relationships through a thirteen-dimensional constellation

Enhanced recognition accuracy in multi-speaker environments

Abstract

This paper proposes an original statistical decision theory to accomplish a multi-speaker recognition task in cocktail party problem. This theory relies on an assumption that the varied frequencies of speakers obey Gaussian distribution and the relationship of their voiceprints can be represented by Euclidean distance vectors. This paper uses Mel-Frequency Cepstral Coefficients to extract the feature of a voice in judging whether a speaker is included in a multi-speaker environment and distinguish who the speaker should be. Finally, a thirteen-dimension constellation drawing is established by mapping from Manhattan distances of speakers in order to take a thorough consideration about gross influential factors.