Speaker segmentation using deep speaker vectors for fast speaker change scenarios

Wang, Renyu; Gu, Minghong; Li, Lantian; Xu, Mingxing; Zheng, Thomas Fang

doi:10.1109/icassp.2017.7953192

Cited by 16 publications

(10 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Most of the DNN based SD systems introduced in Section 1 use DNN to describe a speaker in a relatively short segment of conversation and then compare two representations of adjacent segments (e.g. so called d-vectors [12]) to decide if the speaker change occurred. On the contrary, our approach using the CNN-based SCD finds the possible speaker changes in spectogram and additionally uses the information for the refinement of accumulation process of statistics.…”

Section: Discussionmentioning

confidence: 99%

“…The speaker change detection (SCD) is often applied to the audio signal to obtain segments which ideally contain a speech of a single speaker [2]. Commonly used approaches to the SCD include the Bayesian Information Criterion (BIC), Generalized Likelihood Ratio (GLR), Kullback-Leibler divergence [8,9], Support Vector Machine (SVM) [10] and Deep Neural Networks (DNNs) [11,12]. However, in a spontaneous telephone conversation containing very short speaker turns and frequent overlapping speech, diarization systems often omit the SCD process and use a simple constant length window segmentation of speech [3,5].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Speaker Diarization Using Convolutional Neural Network for Statistics Accumulation Refinement

2017

View full text Add to dashboard Cite

The aim of this paper is to investigate the benefit of information from a speaker change detection system based on Convolutional Neural Network (CNN) when applied to the process of accumulation of statistics for an i-vector generation. The investigation is carried out on the problem of diarization. In our system, the output of the CNN is a probability value of a speaker change in a conversation for a given time segment. According to this probability, we cut the conversation into short segments that are then represented by the i-vector (to describe a speaker in it). We propose a technique to utilize the information from the CNN for the weighting of the acoustic data in a segment to refine the statistics accumulation process. This technique enables us to represent the speaker better in the final i-vector. The experiments on the English part of the CallHome corpus show that our proposed refinement of the statistics accumulation is beneficial with the relative improvement of Diarization Error Rate almost by 16 % when compared to the speaker diarization system without statistics refinement.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Speaker Diarization Using Convolutional Neural Network for Statistics Accumulation Refinement

2017

View full text Add to dashboard Cite

show abstract

“…Alternatively, deep neural networks (DNNs) have also been successfully utilized to extract complex features [7,8]. Furthermore, d-vectors were presented in [9], yielding excellent results. The latest trend goes in the direction of deep speaker embeddings [10,11] designed for endto-end systems.…”

Section: Related Workmentioning

confidence: 99%

An Approach to Online Speaker Change Point Detection Using DNNs and WFSTs

2019

View full text Add to dashboard Cite

In this paper, a new approach to speaker change point (SCP) detection is presented. This method is suitable for online applications (e.g., real-time broadcast monitoring). It is designed in a series of consecutive experiments, aiming at quality of detection as well as low latency. The resulting scheme utilizes a convolution neural network (CNN), whose output is smoothed by a decoder. The CNN is trained using data complemented by artificial examples to reduce different types of errors, and the decoder is based on a weighted finite state transducer (WFST) with the forced length of the transition model. Results obtained on data taken from the COST278 database show that our online approach yields results comparable with an offline multi-pass LIUM toolkit while operating online with a low latency.

show abstract

“…After extracting and segmenting audio recording into frames with 25ms length, frames are processed by VAD method introduced in [24] to classify whether the frame is speech or not. We use speech segment with 0.1s window length proposed in [25] to split frames into homogeneous segments discriminately, and then feed these segments into deep recurrent convolutional neural network proposed in [17] to learn speaker embedding.…”

Section: System Architecturementioning

confidence: 99%

A Study of Classifying Style of Teachers a State of Students' Learning based on K12 Online Education

Gong¹,

Zheng²

2018

Proceedings of the 2018 3rd International Conference on Education, Sports, Arts and Management Engineering (ICESAME 2018)

View full text Add to dashboard Cite

Abstract. In recent years, online education has been advancing significantly. However there is a major challenge how to evaluate style of teachers and state of student learning. In this paper, we propose a novel method that combines speaker diarization, speaker recognition, feature selection to classify style of teachers and state of students learning based on audio data. We train speaker recognition model and learn embedding vector of teachers or students on online platform. We select 25 acoustic features and statistical features from audio recordings and train classification model to classify style of teachers and state of students' learning jointly. Experimental results show that the task of classifying style of teachers achieves 71.25% precision and precision of classifying state of students' learning is 83.71%.

show abstract

Speaker segmentation using deep speaker vectors for fast speaker change scenarios

Cited by 16 publications

References 7 publications

Speaker Diarization Using Convolutional Neural Network for Statistics Accumulation Refinement

Speaker Diarization Using Convolutional Neural Network for Statistics Accumulation Refinement

An Approach to Online Speaker Change Point Detection Using DNNs and WFSTs

A Study of Classifying Style of Teachers a State of Students' Learning based on K12 Online Education

Contact Info

Product

Resources

About