Text independent speaker recognition system using GMM

Bagul, Sudhir; Shastri, Rajveer

doi:10.1109/ichci-ieee.2013.6887781

Cited by 8 publications

(2 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This is because the GMM model is characterized by the parameters related averages and variance of data also allow modeling of data distribution with optional precision. In the same way, GMM proved appropriate for the problem of recognizing information contours such as speaker recognition, dialect recognition, language identification, emotion recognition, and music genre identification [61][62][63][64][65][66][67]. On the other hand, in terms of model implementation, GMM allows for training in a much shorter time than ANN, leading to unnecessary use of complex and expensive hardware configurations including GPUs.…”

Section: Gaussian Mixture Modelmentioning

confidence: 99%

Some New Results on Automatic Identification of Vietnamese Folk Songs Cheo and Quanho

Thành¹,

Loan²,

Quang³

2020

JCC

View full text Add to dashboard Cite

Vietnamese folk songs are very rich in genre and content. Identifying Vietnamese folk tunes will contribute to the storage and search for information about these tunes automatically. The paper will present an overview of the classification of music genres that have been performed in Vietnam and abroad. For two types of very popular folk songs of Vietnam such as Cheo and Quan ho, the paper describes the dataset and GMM (Gaussian Mixture Model) to perform the experiments on identifying some of these folk songs. The GMM used for experiment with 4 sets of parameters containing MFCC (Mel Frequency Cepstral Coefficients), energy, first derivative and second derivative of MFCC and energy, tempo, intensity, and fundamental frequency. The results showed that the parameters added to the MFCCs contributed significantly to the improvement of the identification accuracy with the appropriate values of Gaussian component number M. Our experiments also showed that, on average, the length of the excerpts was only 29.63% of the whole song for Cheo and 38.1% of the whole song for Quan ho, the identification rate was only 3.1% and 2.33% less than the whole song for Cheo and Quan ho respectively.

show abstract

Section: Gaussian Mixture Modelmentioning

confidence: 99%

Some New Results on Automatic Identification of Vietnamese Folk Songs Cheo and Quanho

Thành¹,

Loan²,

Quang³

2020

JCC

View full text Add to dashboard Cite

show abstract

“…It comes in two forms that speaker verification [7] aims to verify if two audio samples belong to the same speaker, whereas speaker identification [58] predicts the identity of the speaker given an audio sample. Speaker recognition has been approached by applying a variety of machine learning models [4,5,12], either standard or specifically designed, to speech features such as MFCC [34]. For many years, the GMM-UBM framework [44] dominated this particular field.…”

Section: Audio-based Speaker Identificationmentioning

confidence: 99%

iQIYI-VID: A Large Dataset for Multi-modal Person Identification

Liu,

Peng,

Shi

et al. 2018

Preprint

View full text Add to dashboard Cite

Person identification in the wild is very challenging due to great variation in poses, face quality, clothes, makeup and so on. Traditional research, such as face recognition, person re-identification, and speaker recognition, often focuses on a single modal of information, which is inadequate to handle all the situations in practice. Multimodal person identification is a more promising way that we can jointly utilize face, head, body, audio features, and so on. In this paper, we introduce iQIYI-VID, the largest video dataset for multi-modal person identification. It is composed of 600K video clips of 5,000 celebrities. These video clips are extracted from 400K hours of online videos of various types, ranging from movies, variety shows, TV series, to news broadcasting. All video clips pass through a careful human annotation process, and the error rate of labels is lower than 0.2%. We evaluated the state-of-art models of face recognition, person reidentification, and speaker recognition on the iQIYI-VID dataset. Experimental results show that these models are still far from being perfect for the task of person identification in the wild. We proposed a Multi-modal Attention module to fuse multi-modal features that can improve person identification considerably. We have released the dataset online to promote multi-modal person identification research.

show abstract