Efficient Speaker Change Detection Using Adapted Gaussian Mixture Models

Abstract

A new approach to speaker change detection is proposed and investigated. The method, which is based on a probabilistic framework, provides an effective means for tackling the problem posed by phonetic variation in high-resolution speaker change detection. Additionally, the approach incorporates the capability for dealing with undesired effects of variations in speech characteristics. Using the experimental investigations conduced with clean and broadcast news audio, it is shown that the proposed method is significantly more effective than the currently popular techniques for speaker change detection. To enhance the computational efficiency of the proposed method, modified implementation algorithms are introduced which are based on the exploitation of the redundant operations and a fast scoring procedure. It is shown that, through the use of the proposed fast algorithm, the computational efficiency of the approach can be increased by over 77% without significant reduction in its accuracy. The paper discusses the principles and characteristics of the proposed speaker change detection method, and provides a detailed description of its efficient implementation. The experiments, investigating the performance of the proposed method and its effectiveness in relation to other approaches, are described and an analysis of the results is presented.

Speaker Change Detection Based on a Graph-Partitioning Criterion

“…To evaluate the proposed method, we tested our method on the conTIMIT dataset 1) in [7] and the speech-excerpt dataset from movies, which is constructed by the authors and available In evaluating SCD performance, we use the false alarm rate (FAR) and the miss detection…”

“…This paper focuses on one of the issues, speaker change detection (SCD) of an audio stream. SCD is an indispensable prior step for speaker segmentation and clustering, which has a wide range of applications including audio indexing and retrieval, speaker tracking and identification, and movie summarization [1].…”

“…A comparative detailed review of the previous SCD methods is given in [2]. Among the various metrics, the generalized likelihood ratio (GLR) based on the Bayesian information criterion (BIC) [5][6][7] is considered as a standard approach for SCD, which involves a statistical hypothesis test between the null hypothesis (no speaker change in the analysis window) and the alter- before comparing each other, which is known to be vulnerable to phonetic variations especially for a short segment size [1]. In this paper, we seek a different approach using the pairwise The preliminary work on the SCD based on the pairwise distance matrix was presented at [8].…”

Combining synchronous averaging with a Gaussian mixture model novelty detection scheme for vibration-based condition monitoring of a gearbox

“…A related problem is that of speaker change detection, where the problem is to determine the time instances in an audio stream when a voice or speaker change. In recent years GMM-based change detection methodologies have served as the dominant approach for speaker change detection, primarily due to GMM's good ability to detect various acoustic changes [11]. This paper proposes a novel gear monitoring technique, which combines ideas from residual analysis, novelty detection and synchronous averaging.…”

BIC-Based Speaker Segmentation Using Divide-and-Conquer Strategies With Application to Speaker Diarization

“…Window-growing-based segmentation (WinGrow) [11], [15], [23], [17], fixed-size sliding window segmentation (FixSlid) [10], [12], [24], [25], [26], and DISTBIC [9] are three popular distance-based segmentation approaches.…”