Beiqian Dai

University of Science and Technology of China

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Trends

Author Order

Document Type

Co-Authors

Yanlu Xie

Beijing Language and Culture University

Zhiqiang Yao

Anhui University

Xi Zhou

Guangdong Academy of Sciences

Jun Sun

West China Medical Center of Sichuan University

M. Liu

University of Science and Technology of China

Changfu Wang

Henan University

Wei Lu

Beijing Microelectronics Technology Institute

Hui Li

University of Science and Technology of China

Frank K. Soong

Microsoft Research Asia (China)

Xi Zhou

Institute of Optics and Electronics, Chinese Academy of Sciences

Cooperative Institutions

University of Science and Technology of China

Chinese Academy of Sciences

University of Chinese Academy of Sciences

Shanghai Jiao Tong University

Harbin Institute of Technology

Northwestern Polytechnical University

Nanjing University of Science and Technology

Tsinghua University

Microsoft Research Asia (China)

Microsoft Research (United Kingdom)

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Field

Adaptive recognition method based on posterior use of distribution pattern of output probabilities

4th International Conference on Spoken Language Processing (ICSLP 1996) (1996)

Jinsong Zhang Beiqian Dai Changfu Wang Hingkeung Kwan Keikichi Hirose

10.21437/icslp.1996-251

Cite

Citations (0)

Maximum Likelihood Discriminant Feature for Text-Independent Speaker Verification

Qingsong Liu Beiqian Dai

Feature extraction is an essential first step in speaker verification applications. In addition to static features extracted from each frame of speech data, it is beneficial to use dynamic features that use information from neighboring frames. In this paper a new feature estimation method based on maximum likelihood discriminant analysis is presented. We compare it to traditional MFCC features in a NIST 2006 SRE core task. Experiments show that the proposed scheme provides more discriminative feature vectors. The features obtained with the new estimation method show a 10% -15% relative improvement in EER and MinDCF over traditional MFCC features.

Discriminative model

NIST

Mel-frequency cepstrum

Feature (linguistics)

10.1109/cisp.2009.5301537

Cite

Citations (0)

Kurtosis Normalization in Feature Space for Robust Speaker Verification

Yanlu Xie Beiqian Dai Zhiqiang Yao M. Liu

The acoustic mismatch between the training and test environments will lead to the difference of the statistical characteristics of speech parameters. Since the statistical characteristics of the kurtosis can measure the non-Gaussianity of a random variable, kurtosis normalization will make the training and test speech parameters match the standard normal distribution in some sense. In this paper, a kurtosis normalization method using sigmoid functions (logit functions) in feature space is presented for GMM-UBM based text-independent speaker verification system. Experimental results on the 2004 NIST SRE database show that with the new method significant improvement can be achieved in not only equal error rate but also minimum detection cost compared with baseline system (more than 33% relative reduction for long speech).

Kurtosis

Normalization

NIST

Word error rate

Sigmoid function

Speaker Verification

Feature vector

10.1109/icassp.2006.1659971

Cite

Citations (6)

A Pitch Detection Algorithm Based on AMDF and ACF

Hui Li Beiqian Dai Wei Lu

The paper proposes a pitch detection algorithm based on the short-time average magnitude difference function (AMDF) and the short-term autocorrelation function (ACF). At first, AMDF values are computed by AMDF algorithm for a frame of speech signal. And then ACF values are computed by ACF algorithm for the AMDF values. In order to decreases computational expense and complexity, the AMDF values of the frame of speech signal are then transformed into one bit signals. The method can also decrease the effects of amplitude and formants the speech signal for pitch detection. The pitch period is calculated by ACF algorithm for the one bit signals. The multiplication operation for short-time autocorrelation function of the one bit signals is replaced by simple addition operation. The algorithm can largely reduce its computational cost compared with calculating ACF directly. The experiments show that the algorithm can decreases computational expense and complexity compared with ACF algorithm. The algorithm can decrease the error rate of pitch detection compared with AMDF algorithm. The proposed algorithm is easy to implement in real-time signal processing

Pitch detection algorithm

SIGNAL (programming language)

10.1109/icassp.2006.1660036

Cite

Citations (57)

Directly Modeling of Correlation Matrices for GMM in Speaker Identification

2022 26th International Conference on Pattern Recognition (ICPR) (2006)

Zhiqiang Yao Xi Zhou Beiqian Dai M. Liu Yanlu Xie

In this paper, we present a new framework to model full covariance matrices of Gaussian components. In this framework, directly modeling the full correlation matrix instead of the full covariance matrix is our purpose, as the correlation matrix is the direct description of the correlation of inter-feature elements. In order to model full correlation matrices, we share linear transformations among components' full correlation matrices. Thus, the full correlation matrix of each component is represented by a shared linear transformation and a componentspecific diagonal correlation matrix. The transformation is used to help the diagonal correlation matrix to model the correlation of inter feature-vector elements more precisely. We evaluate our new framework on a mandarin speaker identification task. Experiments show that above 35% reduction in speaker identification error rate is achieved compared with the best diagonal covariance models. Furthermore, our algorithm achieved better performance than STC does.

Matrix (chemical analysis)

Transformation matrix

Canonical correlation

Diagonal matrix

10.1109/icpr.2006.453

Cite

Citations (0)

Improving speaker verification with figure of merit training

IEEE International Conference on Acoustics Speech and Signal Processing (2002)

Xiaohan Li Chang Chang Beiqian Dai

Speaker Verification

Figure of Merit

10.1109/icassp.2002.1005834

Cite

Citations (1)

Glottal Flow Derivative Conversion from Source Speaker to Target Speaker Based on Legendre Orthogonal Decomposition

Journal of Chinese Computer Systems (2009)

Beiqian Dai

An automated glottal waveform estimation algorithm which does not need to estimate the precise glottal closure information(GCI)is used.The algorithm uses only basic approximations of glottal closure regions and successive iterations to find the best candidate for a glottal waveform estimate within a speech frame.To parameterize the glottal flow derivative,we propose a novel method which employs the n-dimensional Legendre orthogonal coefficients to form a vector describing the shape of glottal flow derivative.This method eliminates the relativity among each dimension within the vector.More over,Euclidian distance between two vectors just reveals the similarity of their shapes,thus performance of proposed method is better than normal polynomial coefficients.Compared with normal polynomial coefficients based VQ conversion algorithm,the performance of proposed approach is improved by 11.92%.

Similarity (geometry)

Source

Cite

Citations (0)

Kurtosis Normalization after Short-Time Gaussianization for Robust Speaker Verification

Yanlu Xie Beiqian Dai Jun Sun

Statistical normalization has been popularly used to transform the distribution of the speech feature to decrease the mismatch between the training and test environment in robust speaker verification. This paper presents a method to normalization kurtosis which can measure the non-Gaussianity in quantity. The kurtosis normalization method proposed here is based on the cumulative distribution function (CDF) of a sub-Gaussian function and used after the short-time Gaussianization normalization method in feature space. Thus the distribution of speech parameters becomes more approximate the normal one. Experimental results based on the 2001 NIST SRE database show that when the short-time non-Gaussianity and kurtosis are both normalized, significant improvement could be achieved in not only equal error rate but also minimum detection cost compared to baseline system

Kurtosis

Normalization

NIST

Speaker Verification

10.1109/wcica.2006.1713834

Cite

Citations (3)

A New Method of Score Normalization for Text-Independent Speaker Verification

Signal Processing (2006)

Beiqian Dai

Normalization

Speaker Verification

Standard score

Source

Cite

Citations (0)

Chinese speech understanding system

Xiaodong Wang Beiqian Dai

Application of the knowledge of syntax, morphology, and semantics to acoustic analysis and semantic achievement is discussed. An acoustic analyzer which is restricted by syntax rules and morphology rules is proposed. The semantic achievement procedure based on syntactic structure analysis and semantic deep structure is used to obtain the meaning of input speech. In addition, to make the system accept fairly natural speech, a multiparameter segmentation method is described. The authors present experimental results obtained using the system.< >

10.1109/icpr.1988.28448

Cite

Citations (1)