음성 부호화 음질 측정 방법

서정태 한국교통대학교 2020 한국교통대학교 논문집 Vol.55 No.-

This paper presents a robust speech quality measure method for decoded speech. Speech quality is an important means of estimating performance of decoded speech in which most of the transmitted voice signals are corrupted by background noise. Quality information under varying noise environments has traditionally been determined using subjective tests based on human listener rating schemes such as MOS(Mean Opinion Score) and the tests are quite expensive and highly dependent on the test conditions. As a result, it is required to develop an speech quality measure method that can provide a good estimate of subjective quality using the easily obtained objective measure. There has been few studies on practical applications of the real-time system to the mobile system area and especially any channel distortion problems due to the connected local wire-line. In this paper, we study speech quality measure method. From the experimental results, our measure method showed good performance.

버스트 손실 시 부호화기의 음질개선

서정태 忠州大學校 2009 한국교통대학교 논문집 Vol.44 No.-

We have been used widely for a Low Bit Rate Speech Coder. G.729 Speech Coder adopt the advantage of the source codec with the waveform codec. We have the Low Bit Rate with superior tone as the compositeness Speech Coder. It does the analysis law by the synthesis basic to the foundation. It is superior we making the complex sound of the best suited we operating so that we produce the parameter. Research accomplished the quality of sound change of the original pronunciation and time which used Speech Coder. The criteria of the Speech Coder assessment used FER(Frame Erasure Rate), BER(Bit Error Rate) to become and the quality of sound assessment was a planning low. We became a quality assessment criteria and used MOS(Mean Opinion Score) measurement. We measure the voice quality and evaluate the performance enhance of the tone the comparison analyzes the quality of sound according to the error rate.

음성부호화기[ITU-T (8Kbps)] 알고리듬 분석

徐廷泰 충주대 2001 한국교통대학교 논문집 Vol.36 No.2

In this paper, major subject contains the analysis of an algorithms for the coding of speech signals at 8kbps using Conjugate-Structure Algebraic-Code-Excited Linear-Prediction(CS-ACELP), or Recommendation ITU-T (G.729). This speech coder is designed to operate with a digital signal obtained by first performing PSTN(Public Switched Telephone Network) bandwidth filtering of the analog input signal, then sampling it at 8KHz, followed by conversion to 16-bit linear PCM for the input to the encoder. This paper described a toll quality medium-delay 8kbps speech coder.This coder(ITU-T[G.729]) can be used for a variety of applications.In terms of its inherent robustness against channel errors and the provision for the concealment of detected frame erasures it has a strong potential for wireless applications.

음성 부호화기 LPC계수 예측 알고리듬 연구

서정태 韓國交通大學校 2013 한국교통대학교 논문집 Vol.48 No.-

The most important information in human communication is speech. However, since it requires a large number of bits to express speech in digital form, it is necessary to develop a high-performance speech coding algorithm for low bit rate which uses a limited channel capacity . The researches and developments on the low bit rate speech coder have focused on the quality of synthesized sound. A dominant factor that determines the sound quality is the choice of how to model speech signal and how to estimate relevant parameters. The degradation of synthesized sound is caused by the speech generation model itself and inaccuracy in modelling parameters. In this paper, it is focused on search algorithm formant factor of LPC coefficient.

음성 부호화기[V-SELP]장 구간 예측 연구

서정태 한국교통대학교 2014 한국교통대학교 논문집 Vol.49 No.-

Code Excited Linear Prediction(celp) Speech Coders exhibit good performance at data rate as low as 4.8Kbps. The major drawback to CELP type Coders is their large computational burden. The vector sum excited linear prediction speech coder utilizes a codebook with a structure which allows for a very efficient search procedure. the code book is organized with a predefined structure which signifucantly reduces the time required for the optimum code word search. v-selp codec utilizes three excitation sources. The first is from long term predictor state, the scond and third sources are from two v-selp excitation codebook. In this paper the vector sum excited linear prediction speech coder is presented. It utilizes a codebook with a structure that allows for a very efficient search procedure. Other advantages of the v-selp codebook structure are discussed, and specially detailed study of Long-term prediction procedure.

서브밴드를 이용한 넓은 대역 음성 부호화기

徐廷泰 충주대 2004 産業科學論文集 Vol.12 No.-

This paper describes a Sub Band coding scheme for wideband speech signal. In this analyzed coder, band splitting approach is used. The lower band speech is encoded with ITU-T G.729E, which is known as a high-quality narrow band speech coder at 8/11.8 kbps. Since the encoding of the lower band is independent of upper band signal, the output speech of the decoder can be selected to be narrowband or wideband according to the channel conditions. Possible wideband applications include high-quality audio conferencing, digital AM broadcasting and third-generation wireless communications

음성 부호화기의 장 구간 예측 연구

서정태 忠州大學校 2010 한국교통대학교 논문집 Vol.45 No.-

speech is the most important information in human communication. However, since it requires a large number of bits to express speech in digital form, it is essential to develop a high-performance speech coding algorithm for low bit rate which uses a limited channel capacity effectively. The researches and developments on the low bit rate speech coder have focused on the quality of synthesized sound. A dominant factor that determines the sound quality is the choice of how to model speech signal and how to estimate relevant parameters. The degradation of synthesized sound is caused by the speech generation model itself and inaccuracy in modelling parameters. In particular, the quality in low bit rate is known to depend mainly on pitch information. In this paper, a methods improving performance in low bit rate is researched which adopts the LTP(Long Term Predictor) Anaysis on the CELP(Code Excited Linear Predictor)-style speech coder, the most remarkable coding method which provides comparatively good quality in low bit rate.

다대역 다중펄스를 이용한 낮은 비트율 음성 부호화기

서정태 충주대 산업과학기술연구소 1996 産業科學論文集 Vol.4 No.-

In this paper, a methods improving prformance in low bit rate under 4.8kbps is proposed which adopts the secondary LTP(Long Term Predictor) on the CELP(Code Excited Linear Predictor style speech coder. Conventional CELP speech coder, quasi periodic components are still remained after LTP. The proposed methods improves the quality of Synthesized sound by eliminating the quasi periodicty effectively. It is done by dividing the residual signnal to a number of bands and subsequently eliminating once more with multi-pulse modelled to correspond with characteristics of residual signal.

적은 전송 율에서의 예측기와 단구간 추정

서정태 韓國交通大學校 2022 한국교통대학교 논문집 Vol.57 No.-

The important information in communication is human’s speech. But it requires a huge bits to express speech in digital. Prediction and STP(Short Term Prediction) is a method used mostly in speech processing for representing the spectral envelope of a digital signal of speech in compressed form, using the information of a linear predictive model. It is the most widely used method in speech coding and speech synthesis. It is a powerful speech analysis technique, and a useful method for encoding good quality speech at a low bit rate. It starts with the assumption that a speech signal is produced by a buzzer at the end of a tube (for voiced sounds), with occasional added hissing and popping sounds (for voiceless sounds such as sibilants and plosives). Although apparently crude, this Source–filter model is actually a close approximation of the reality of speech production. In this paper, understanding of speech coder structure and predictor is performed.

음성 신호를 사용한 GMM기반의 감정 인식

서정태,김원구,강면구 한국음향학회 2004 韓國音響學會誌 Vol.23 No.3

본 논문은 화자 및 문장 독립적 감정 인식을 위한 특징 파라메터와 패턴인식 알고리즘에 관하여 연구하였다. 본 논문에서는 기존 감정 인식 방법과의 비교를 위하여 KNN을 이용한 알고리즘을 사용하였고, 화자 및 문장 독립적 감정 인식을 위하여 VQ와 GMM을 이용한 알고리즘을 사용하였다. 그리고 특징으로 사용한 음성 파라메터로 피치, 에너지, MFCC, 그리고 그것들의 1, 2차 미분을 사용하였다. 실험을 통해 피치와 에너지 파라메터를 사용하였을 때보다 MFCC와 그 미분들을 특징 파라메터로 사용하였을 때 더 좋은 감정 인식 성능을 보였으며, KNN과 VQ보다 GMM을 기반으로 한 인식 알고리즘이 화자 및 문장 독립적 감정 인식 시스템에서 보다 적합하였다. This paper studied the pattern recognition algorithm and feature parameters for speaker and context independent emotion recognition. In this paper, KNN algorithm was used as the pattern matching technique for comparison, and also VQ and GMM were used for speaker and context independent recognition. The speech parameters used as the feature are pitch. energy, MFCC and their first and second derivatives. Experimental results showed that emotion recognizer using MFCC and its derivatives showed better performance than that using the pitch and energy parameters. For pattern recognition algorithm. GMM-based emotion recognizer was superior to KNN and VQ-based recognizer.