Speaker Identification Based on Log Area Ratio and Gaussian Mixture Models in Narrow-Band Speech

Chow, David; Abdulla, Waleed H.

doi:10.1007/978-3-540-28633-2_95

Cited by 26 publications

(18 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [14] the authors emphasize the importance of the appropriate feature set. They have suggested a new feature set called the perceptual log area ratio (PLAR).…”

Section: Related Workmentioning

confidence: 98%

Survey of Automated Speaker Identification Methods

Sidorov

Schmitt

Zablotskiy

et al. 2013

2013 9th International Conference on Intelligent Environments

View full text Add to dashboard Cite

In this paper we present an overview of state-of-the-art approaches for speaker identification. Due to the increased number of dialogue system applications the interest in that field has grown significantly in recent years. Nevertheless, there are many open issues in the field of automatic speaker identification. Among them the choice of the appropriate speech signal features and machine learning algorithms could be mentioned. We make here an overview of modern methods designed for the problem of speaker identification. We also describe here our direction for possible improvements to the automated speaker identification.

show abstract

“…In [14] the authors emphasize the importance of the appropriate feature set. They have suggested a new feature set called the perceptual log area ratio (PLAR).…”

Section: Related Workmentioning

confidence: 98%

Survey of Automated Speaker Identification Methods

Sidorov

Schmitt

Zablotskiy

et al. 2013

2013 9th International Conference on Intelligent Environments

View full text Add to dashboard Cite

show abstract

“…Two common choices are Mel Frequency Cepstral Coefficients (MFCC) and Linear Predictive Coding (LPC). These two features are widely used for speech recognition and speaker identification [3] and [11]. It seems that MFCC is better and more popular than LPC [18].…”

Section: Feature Extractionmentioning

confidence: 99%

Large corpus of Iranian music

Shirali-Shahreza

2009

2009 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT)

View full text Add to dashboard Cite

Digital music is now widely used and a great number of digital music files are available on the Internet. Designing systems that can automatically index, search and retrieve the digital music is one of the active research fields. One of the requirements for design and evaluation of such systems is corpuses of digital music samples. A good corpus is expected to have a number of features such as including a great number of samples, have samples of different artists and also samples with different quality. Additionally, the regional music of different parts of the world uses local musical instruments which results in music that are acoustically different. So, corpuses of regional music are needed for design and evaluation of web scale systems. Considering the above reasons, we created a large corpus of Iranian music. In this paper, we review available public music corpuses and then describe our corpus. Our corpus contains 27496 music tracks of 1355 artists that gathered from the Internet. The lyrics for more than 54% of tracks are also provided. We analyze the different aspects of collected tracks such as their bit rate and sampling rate which are helpful in designing large scale systems.

show abstract

“…This coding assumes that an acoustic signal is generated by the tube. The acoustic signal can be also characterized by the frequency and intensity [16], [17]. LPC calculates a set of coefficients.…”

Section: Linear Predictive Codingmentioning

confidence: 99%

“…The block calculates the N-th reflection coefficient value using the formula rc N = -a NN for N-th order LPC vector LN=[1, a N1 , a N2 ,…, a NN ]. After that it finds the lower order LPC vectors, LN-1, LN-2,…, L1 [16], [17]. Next the reflection coefficients are obtained [rc 1 , rc 2 , … , rc N ].…”

Section: Reflection Coefficientsmentioning

confidence: 99%

Diagnostics of DC and Induction Motors Based on the Analysis of Acoustic Signals

Głowacz

2014

Measurement Science Review

View full text Add to dashboard Cite

In this paper, a non-invasive method of early fault diagnostics of electric motors was proposed. This method uses acoustic signals generated by electric motors. Essential features were extracted from acoustic signals of motors. A plan of study of acoustic signals of electric motors was proposed. Researches were carried out for faultless induction motor, induction motor with one faulty rotor bar, induction motor with two faulty rotor bars and flawless Direct Current, and Direct Current motor with shorted rotor coils. Researches were carried out for methods of signal processing: log area ratio coefficients, Multiple signal classification, Nearest Neighbor classifier and the Bayes classifier. A pattern creation process was carried out using 40 samples of sound. In the identification process 130 five-second test samples were used. The proposed approach will also reduce the costs of maintenance and the number of faulty motors in the industry.

show abstract

Speaker Identification Based on Log Area Ratio and Gaussian Mixture Models in Narrow-Band Speech

Cited by 26 publications

References 8 publications

Survey of Automated Speaker Identification Methods

Survey of Automated Speaker Identification Methods

Large corpus of Iranian music

Diagnostics of DC and Induction Motors Based on the Analysis of Acoustic Signals

Contact Info

Product

Resources

About