Binary quantization of feature vectors for robust text-independent speaker identification

Yuan, Zhong-Xuan; Xu, Bo; Yu, Chongzhi

doi:10.1109/89.736332

Cited by 20 publications

(6 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In general, textdependent systems are more reliable and accurate, since both the content and voice can be compared [3], [4]. Speaker Recognition systems have been developed for a wide range of applications [6] - [9]. Although many new techniques have been developed, widespread deployment of applications and services is still not possible.…”

Section: Introductionmentioning

confidence: 98%

Automatic speaker recognition using circular DFT sectors

Kekre

Kulkarni

2011

Proceedings of the International Conference &Amp; Workshop on Emerging Trends in Technology - ICWET '11

View full text Add to dashboard Cite

In this paper, we propose automatic speaker recognition using circular DFT (Discrete Fourier Transform) sectors. In the first method, the feature vectors are extracted by dividing the complex DFT spectrum into circular sectors and then taking the weighted density count of the number of points in each of these sectors. In the second and third method, the circular sectors are further divided and then weighted density count of the number of samples is used as feature vector. The results show that this approach gives fairly good speaker recognition (70 % -80%)Also the results improve as the circular sectors are further divided.

show abstract

Section: Introductionmentioning

confidence: 98%

Automatic speaker recognition using circular DFT sectors

Kekre

Kulkarni

2011

Proceedings of the International Conference &Amp; Workshop on Emerging Trends in Technology - ICWET '11

View full text Add to dashboard Cite

show abstract

“…For instance, it has been shown that social network users' relationships can be used to discover their identities and geo-locations and even track them [39,43]. Some other application examples involving binary-and matrix-valued data include the coarse quantization for data compression and storage [6,17,65], and the recently emerged XNOR-Nets [54] (where the queried data is the binarized training images). As a result, designing a differentially private mechanism for binary-and matrix-valued queries is in dire need.…”

Section: Introductionmentioning

confidence: 99%

Differentially private binary- and matrix-valued data query

Yılmaz

et al. 2021

Proc. VLDB Endow.

View full text Add to dashboard Cite

Differential privacy has been widely adopted to release continuous- and scalar-valued information on a database without compromising the privacy of individual data records in it. The problem of querying binary- and matrix-valued information on a database in a differentially private manner has rarely been studied. However, binary- and matrix-valued data are ubiquitous in real-world applications, whose privacy concerns may arise under a variety of circumstances. In this paper, we devise an exclusive or (XOR) mechanism that perturbs binary- and matrix-valued query result by conducting an XOR operation on the query result with calibrated noises attributed to a matrix-valued Bernoulli distribution. We first rigorously analyze the privacy and utility guarantee of the proposed XOR mechanism. Then, to generate the parameters in the matrix-valued Bernoulli distribution, we develop a heuristic approach to minimize the expected square query error rate under ϵ -differential privacy constraint. Additionally, to address the intractability of calculating the probability density function (PDF) of this distribution and efficiently generate samples from it, we adapt an Exact Hamiltonian Monte Carlo based sampling scheme. Finally, we experimentally demonstrate the efficacy of the XOR mechanism by considering binary data classification and social network analysis, all in a differentially private manner. Experiment results show that the XOR mechanism notably outperforms other state-of-the-art differentially private methods in terms of utility (such as classification accuracy and F 1 score), and even achieves comparable utility to the non-private mechanisms.

show abstract

“…All upper diagonal and diagonal values of Kekre transform matrix are one, while the lower diagonal part except the values just below diagonal are zero. Generalized N×N Kekre Transform Matrix can be given as in (11). The formula for generating the term Kxy of Kekre transform matrix is given by (12).…”

Section: (2) (3)mentioning

confidence: 99%