• The Journal of the Acoustical Society of Korea
• /
• v.31 no.4
• /
• pp.266-272
• /
• 2012

Speaker segmentation is a process of automatically detecting the speaker boundary points in the audio data. Speaker segmentation methods are divided into two categories depending on whether they use a prior knowledge or not: One is the model-based segmentation and the other is the metric-based segmentation. In this paper, we introduce the eigenvoice-based speaker weight distance method and compare it with the representative metric-based methods. Also, we employ and compare the Euclidean and cosine similarity functions to calculate the distance between speaker weight vectors. And we verify that the speaker weight distance method is computationally very efficient compared with the method directly using the distance between the speaker adapted models constructed by the eigenvoice technique.

• The Journal of the Acoustical Society of Korea
• /
• v.42 no.6
• /
• pp.536-543
• /
• 2023

Speaker diarization, which labels for "who spoken when?" in speech with multiple speakers, has been studied on a deep neural network-based end-to-end method for labeling on speech overlap and optimization of speaker diarization models. Most deep neural network-based end-to-end speaker diarization systems perform multi-label classification problem that predicts the labels of all speakers spoken in each frame of speech. However, the performance of the multi-label-based model varies greatly depending on what the threshold is set to. In this paper, it is studied a speaker diarization system using single-label classification so that speaker diarization can be performed without thresholds. The proposed model estimate labels from the output of the model by converting speaker labels into a single label. To consider speaker label permutations in the training, the proposed model is used a combination of Permutation Invariant Training (PIT) loss and cross-entropy loss. In addition, how to add the residual connection structures to model is studied for effective learning of speaker diarization models with deep structures. The experiment used the Librispech database to generate and use simulated noise data for two speakers. When compared with the proposed method and baseline model using the Diarization Error Rate (DER) performance the proposed method can be labeling without threshold, and it has improved performance by about 20.7 %.

• Journal of IKEEE
• /
• v.6 no.1 s.10
• /
• pp.54-64
• /
• 2002

In this paper, we propose the algorithm for segmenting the moving speaker and tracking its movement in the video conference image. For real time processing, we simplify the algorithm which is processed in the order of the segmenting and the tracking step. In the segmenting step, the speaker object is segmented from the image by using both the motion information obtained from the difference method and the illuminance information of image. The reference mask image is created from segmented speaker object. In the tracking step, the moving speaker is tracked by using simple block matching algorithm of which computation time is reduced by discarding the blocks which are classified into the unuseful blocks. In the simulation, we can get the good result of segmenting and tracking the moving speaker by applying the proposed algorithm to several test images.

• The KIPS Transactions:PartB
• /
• v.10B no.6
• /
• pp.673-680
• /
• 2003

In this paper, we study the enhancement of VQ (Vector Quantization) design for text independent speaker recognition. In a concrete way, we present a non-iterative method which makes a vector quantization codebook and this method performs non-iterative learning so that the computational complexity is epochally reduced The proposed Classified Space VQ (CSVQ) design method for text Independent speaker recognition is generalized from Semi-noniterative VQ design method for text dependent speaker recognition. CSVQ contrasts with the existing desiEn method which uses the iterative learninE algorithm for every traininE speaker. The characteristics of a CSVQ design is as follows. First, the proposed method performs the non-iterative learning by using a Classified Space Codebook. Second, a quantization region of each speaker is equivalent for the quantization region of a Classified Space Codebook. And the quantization point of each speaker is the optimal point for the statistical distribution of each speaker in a quantization region of a Classified Space Codebook. Third, Classified Space Codebook (CSC) is constructed through Sample Vector Formation Method (CSVQ1, 2) and Hyper-Lattice Formation Method (CSVQ 3). In the numerical experiment, we use the 12th met-cepstrum feature vectors of 10 speakers and compare it with the existing method, changing the codebook size from 16 to 128 for each Classified Space Codebook. The recognition rate of the proposed method is 100% for CSVQ1, 2. It is equal to the recognition rate of the existing method. Therefore the proposed CSVQ design method is, reducing computational complexity and maintaining the recognition rate, new alternative proposal and CSVQ with CSC can be applied to a general purpose recognition.

• The Journal of the Acoustical Society of Korea
• /
• v.17 no.5
• /
• pp.69-72
• /
• 1998

본 논문에서는 문장독립형 화자인식 시스템에 MQ(Matrix Quantization) 방법 사용 을 제안한다. 또한 인식율을 높이기 위해 MQ를 수정한 방법인 SMQ(Separated Matrix Quantization)을 제안한다. 기존의 VQ-distortion 방법은 대체로 좋은 성능을 가지나 화자의 동적 특성을 이용하지 못한다는 단점이 있다. MQ와 SMQ는 화자의 동적 특성을 이용할 수 있으므로 시간 변화에 대한 화자의 특징 변화까지 모델링 할 수 있는 장점이 있다. MQ는 여러 프레임을 묶어 Matrix Codebook을 가지며 SMQ는 MQ의 기본 codebook을 다시 켑스 트럼의 차수에 따라 나누어 codebook을 만든다. 즉, 켑스트럼 차수를 저, 중, 고차로 나누어 각 부분별로 Matrix codebook을 만들도록 한다. 인식실험은 문장독립 음성 데이터에 대해 실행했으며 MQ모델의 경우 Matrix의 크기를 짧은 음소크기부터 음절단위까지 변화시켜 실 험하였다. 아울러 SMQ 모델에서의 실험은 차수별 유용도를 보기 위하여 부분 차수를 이용 하여 실험하였다. 실험결과 MQ와 SMQ방법이 VQ에 비해 좋은 성능을 가짐을 확인하였다.

• Proceedings of the Acoustical Society of Korea Conference
• /
• autumn
• /
• pp.127-130
• /
• 2004

기존의 전대역(Full-Band)에서 특징 파라미터를 추출하는 화자 확인(Speaker Verification) 시스템은 저대역이나 고대역에서 화자 정보의 특징이 제거되기 쉽다. 또한, 주파수 스펙트럼에 부분적으로 오염이 되는 경우, 특징 파라미터를 왜곡시켜 화자 확인 시스템의 성능을 저하시킨다. 본 논문에서는 이러한 문제점을 해결하기 위해 다중대역 공분산 모델(Covariance Model)을 제안한다. 제안한 방법은 주파수 영역에서 전대역을 여러 개의 부대역(Sub-Band)으로 분할하고, 부대역별로 독립적으로 특징 파라미터를 추출하여 공분산 모델을 구한다. 제안된 방법의 성능 확인을 위하여 공분산 모델 간의 거리를 측정하는 화자 확인 실험을 하였다. 잡음 환경에서 기존의 방법인 전대역에 기반한 공분산 모델과 제안한 방법을 비교 분석한 결과, 제안한 방법이 기존 방법보다 $2\%$정도 성능이 향상되었다. 또한, 제안된 방법은 전대역에 기반한 파라미터 차원 수를 다중대역의 개수로 분할하여 사용하므로 계산량의 감소와 저장 공간면에서 효율적이다.

• The Journal of the Acoustical Society of Korea
• /
• v.34 no.4
• /
• pp.303-309
• /
• 2015

Speaker diarization is the task of determining the speakers for unlabeled data, and DBSCAN (Density-Based Spatial Clustering of Applications with Noise) has been widely used in the field of speaker diarization for its simplicity and computational efficiency. One challenging issue, however, is that if different clusters in non-spatial dataset are adjacent to each other, over-clustering may occur which subsequently degrades the performance of DBSCAN. In this paper, we identify the drawbacks of DBSCAN and propose a new density clustering algorithm based on local distribution property around object. Variable density criterions for local density and spreadness of object are used for effective data clustering. We compare the proposed algorithm to DBSCAN in terms of clustering accuracy. Experimental results confirm that the proposed algorithm exhibits higher accuracy than DBSCAN without over-clustering and confirm that the new approach based on local density and object spreadness is efficient.

• 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.

• The Journal of the Acoustical Society of Korea
• /
• v.19 no.3
• /
• pp.45-50
• /
• 2000

This paper proposes a delta energy method based on Parameter Filtering(PF), which is a speech segmentation algorithm for text dependent speaker verification system over telephone line. Our parametric filter bank adopts a variable bandwidth along with a fixed center frequency. Comparing with other methods, the proposed method turns out very robust to channel noise and background noise. Using this method, we segment an utterance into consecutive subword units, and make models using each subword nit. In terms of EER, the speaker verification system based on whole word model represents 6.1%, whereas the speaker verification system based on subword model represents 4.0%, improving about 2% in EER.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.39 no.2
• /
• pp.116-127
• /
• 2002

A new approach to improve the speaker adaptation algorithm by means of the variable number of observation density functions for CDHMM speech recognizer has been proposed. The proposed method uses the observation density function with more than one mixture in each state to represent speech characteristics in detail. The number of mixtures in each state is determined by the number of frames and the determinant of the variance, respectively. The each MAP Parameter is extracted in every mixture determined by these two methods. In addition, the state segmentation method requiring speaker adaptation can segment the adapting speech more Precisely by using speaker-independent model trained from sufficient database as a priori knowledge. And the state duration distribution is used lot adapting the speech duration information owing to speaker's utterance habit and speed. The recognition rate of the proposed methods are significantly higher than that of the conventional method using one mixture in each state.