• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.66 no.7
• /
• pp.1105-1110
• /
• 2017

Speech with various emotions degrades the performance of the speaker recognition system. In this paper, a speaker recognition method using emotional adaptation has been proposed to improve the performance of speaker recognition system using affective speech. For emotional adaptation, emotional speaker model was generated from speaker model without emotion using a small number of training affective speech and speaker adaptation method. Since it is not easy to obtain a sufficient affective speech for training from a speaker, it is very practical to use a small number of affective speeches in a real situation. The proposed method was evaluated using a Korean database containing four emotions. Experimental results show that the proposed method has better performance than conventional methods in speaker verification and speaker recognition.

• Speech Sciences
• /
• v.7 no.2
• /
• pp.177-189
• /
• 2000

This paper investigates how individual speakers' speech can be distinguished using acoustic parameters such as amplitude, pitch, and formant frequencies. Word samples from fifteen male speakers in their 20's in three different regions were recorded in two different modes (i.e., casual and clear speech) in quiet settings, and were analyzed with a Praat macro scrip. In order to determine individual speakers' acoustical values, the total duration of voicing segments was measured in five different timepoints. Results showed that a high correlation coefficient between $F_1\;and\;F_2$ in formant frequency was found among the speakers although there was little correlation coefficient between amplitude and pitch. Statistical grouping shows that individual speakers' voices were not reflected in regional dialects for both casual and clear speech. In addition, the difference of maximum and minimum in amplitude was about 10 dB which indicates a perceptually audible degree. These acoustic data can give some meaningful guidelines for implementing algorithms of speaker identification and speaker verification.

• Journal of the Korea Society of Computer and Information
• /
• v.5 no.3
• /
• pp.91-98
• /
• 2000

In this paper we design implement the speaker verification system for verifying personal identification using voice. Filter bank magnitude was used as a feature parameter and code-book was made using LBG a1gorithm. The code book convert feature parameters into code sequence. The difference between reference pattern and input pattern measures using DTW(Dynamic Time Warping). The similarity measured using DTW and threshold value derived from deviation were used to discriminate impostor from client speaker.

• The Journal of the Acoustical Society of Korea
• /
• v.18 no.1
• /
• pp.86-91
• /
• 1999

This paper presents the text-prompt method to overcome the weakness of text-dependent and text-independent speaker recognition. Enhanced dynamic time warping for speaker recognition algorithm is applied. For the real-time processing, we use a simple algorithm for end-point detection without increasing computational complexity. The test shows that the weighted-cepstrum is most proper for speaker recognition among various speech parameters. As the experimental results of the proposed algorithm for three prompt words, the speaker identification error rate is 0.02%, and when the threshold is set properly, false rejection rate is 1.89%, false acceptance rate is 0.77% and verification total error rate is 0.97% for speaker verification.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.41 no.3
• /
• pp.175-182
• /
• 2004

There is a growing interest in speaker verification, which verifies someone by his/her voices. There are many speaker vitrification algorithms such as HMM and DTW. However, it is impossible to apply these algorithms to memory limited applications because of large number of feature vectors to register or verify users. In this paper we introduces a speaker verification system using SVM, which needs a little memory usage and computation time. Also we proposed hardware architecture for SVM. Experiments were conducted with Korean database which consists of four-digit strings. Although the error rate of SVM is slightly higher than that of HMM, SVM required much less computation time and small model size.

• Phonetics and Speech Sciences
• /
• v.1 no.4
• /
• pp.127-132
• /
• 2009

The feature vectors which are used in conventional speaker recognition (SR) systems may have many correlations between their neighbors. To improve the performance of the SR, many researchers adopted linear transformation method like principal component analysis (PCA). In general, the linear transformation of the feature vectors is based on concatenated form of the static features and their dynamic features. However, the linear transformation which based on both the static features and their dynamic features is more complex than that based on the static features alone due to the high order of the features. To overcome these problems, we propose an efficient method that applies linear transformation and temporal information of the features to reduce complexity and improve the performance in speaker verification (SV). The proposed method first performs a linear transformation by PCA coefficients. The delta parameters for temporal information are then obtained from the transformed features. The proposed method only requires 1/4 in the size of the covariance matrix compared with adding the static and their dynamic features for PCA coefficients. Also, the delta parameters are extracted from the linearly transformed features after the reduction of dimension in the static features. Compared with the PCA and conventional methods in terms of equal error rate (EER) in SV, the proposed method shows better performance while requiring less storage space and complexity.

• The Journal of the Acoustical Society of Korea
• /
• v.38 no.5
• /
• pp.593-600
• /
• 2019

The similarity in tones between speakers can lower the performance of speaker verification. To improve the performance of speaker verification systems, we propose a multi-task learning technique using deep neural network to learn speaker information and age information. Multi-task learning can improve generalization performances, because it helps deep neural networks to prevent hidden layers from overfitting into one task. However, we found in experiments that learning of age information does not work well in the process of learning the deep neural network. In order to improve the learning, we propose a method to dynamically change the objective function weights of speaker identification and age estimation in the learning process. Results show the equal error rate based on RSR2015 evaluation data set, 6.91 % for the speaker verification system without using age information, 6.77 % using age information only, and 4.73 % using age information when weight change technique was applied.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.2
• /
• pp.137-142
• /
• 2020

Many studies based on I-vector have been conducted in a variety of environments, from text-dependent short-utterance to text-independent long-utterance. In this paper, we propose a speaker verification system employing a combination of I-vector with Probabilistic Linear Discriminant Analysis (PLDA) and speaker embedding of Long Short Term Memory (LSTM) with attention mechanism in far-field and noisy environments. The LSTM model's Equal Error Rate (EER) is 15.52 % and the Attention-LSTM model is 8.46 %, improving by 7.06 %. We show that the proposed method solves the problem of the existing extraction process which defines embedding as a heuristic. The EER of the I-vector/PLDA without combining is 6.18 % that shows the best performance. And combined with attention-LSTM based embedding is 2.57 % that is 3.61 % less than the baseline system, and which improves performance by 58.41 %.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.5
• /
• pp.497-504
• /
• 2020

Constructing speaker embeddings in speaker verification is an important issue. In general, a self-attention mechanism has been applied for speaker embedding encoding. Previous studies focused on training the self-attention in a high-level layer, such as the last pooling layer. In this case, the effect of low-level layers is not well represented in the speaker embedding encoding. In this study, we propose Masked Cross Self-Attentive Encoding (MCSAE) using ResNet. It focuses on training the features of both high-level and low-level layers. Based on multi-layer aggregation, the output features of each residual layer are used for the MCSAE. In the MCSAE, the interdependence of each input features is trained by cross self-attention module. A random masking regularization module is also applied to prevent overfitting problem. The MCSAE enhances the weight of frames representing the speaker information. Then, the output features are concatenated and encoded in the speaker embedding. Therefore, a more informative speaker embedding is encoded by using the MCSAE. The experimental results showed an equal error rate of 2.63 % using the VoxCeleb1 evaluation dataset. It improved performance compared with the previous self-attentive encoding and state-of-the-art methods.

• MALSORI
• /
• no.43
• /
• pp.45-55
• /
• 2002

High reliability of automatic speaker recognition regardless of voice conditions is necessary for forensic application. Audio recordings in real cases are not consistent in voice conditions, such as duration, time interval of recording, given text or conversational speech, transmission channel, etc. In this study the variation of verification error rate of ASR system with the voice conditions was investigated. As a result in order to decrease both false rejection rate and false acception rate, the various voices should be used for training and the duration of train voices should be longer than the test voices.