Disertacijoje nagrinėjama koduoto balso kokybės vertinimo problematika. Pagrindinis dėmesys skiriamas balso kokybės tyrimams, kai perduodama koduota šneka ir prarandami balso paketai. Darbo tikslas yra patobulinti koduoto balso kokybės vertinimo algoritmus. Darbo uždaviniai yra šie: sukurti matavimo priemonę trumpų balso signalo atkarpų kokybei vertinti; apibrėžti koduoto balso segmentų vertės sampratą ir parinkti vertės metrikas; išmatuoti bendrinės šnekos balso segmentų verčių skirstinius; nustatyti skirtingų koderių sukuriamų iškraipymų ribas; ištirti paplitusių koderių inertiškumą, nustatyti kiek laiko pastebima prarastų paketų įtaka sekantiems segmentams. Disertaciją sudaro įvadas, keturi tiriamieji skyriai ir bendrosios išvados. Įvade pristatomas darbo naujumas, aktualumas, aptariamas autoriaus indėlis, formuluojami darbo tikslai. Pirmas skyrius yra apžvalginis – analizuojami balso kokybės vertinimo metodai, jų privalumai ir trūkumai. Kaip savarankiška dalis čia pristatyti autoriaus sudaryti sąrašai lietuviškų žodžių, skirtų šnekos suprantamumo tyrimams. Antrame skyriuje parodoma, kaip galima išplėsti kokybės vertinimo PESQ (angl. Perceptual Evaluation of Speech Quality) algoritmo taikymo ribas. Čia įvedama koduoto balso paketo vertės sąvoka, nustatomi statistiniai paketų vertės skirstiniai. Trečiame skyriuje nagrinėjami specifiniai koduotos šnekos iškraipymai ir kodavimo parametrų įtaka balso kokybei. Parodoma, kad kodavimo iškraipymų dydis priklauso nuo šnekos... [toliau žr. visą tekstą] / The dissertation investigates the problem of quality of coded voice. The main attention is paid to voice quality evaluation under packet loss conditions. The aim of the work is to improve voice quality evaluation algorithms. The tasks of the work are: construction of the means for measurement of voice quality of short voice signals; to define the concept of value of coded voice segment and to choose corresponding value metrics; to measure distributions of frame values in standard voice; to establish limits of distortions created by different codecs; to investigate inertia of wide spread codecs and establish the length of impact of one lost frame. The dissertation consists of the introduction, 4 chapters, conclusions, list of literature. Introduction presents the novelty and topicality of the work, tasks and aims of the work are formulated. The first chapter is overview of voice quality evaluation methods, pros and cons of these methods are analyzed. PESQ algorithm and limits of its applicability are introduced in this chapter too. The lists of Lithuanian words for word intelligibility testing are created. Chapter two presents the method of signal construction that allows to extend PESQ applicability to short signals. This chapter introduces the concept of frame value. Distributions of frame values are calculated. Third chapter analyses distortions created by coding. It is shown that coding distortions depends highly on the signal used and limits of distortion variability are... [to full text]

The dissertation investigates the problem of quality of coded voice. The main attention is paid to voice quality evaluation under packet loss conditions. The aim of the work is to improve voice quality evaluation algorithms. The tasks of the work are: • construction of the means for measurement of voice quality of short voice signals; • to define the concept of value of coded voice segment and to choose corresponding value metrics; • to measure distributions of frame values in standard voice; • to establish limits of distortions created by different codecs; • to investigate inertia of wide spread codecs and establish the length of impact of one lost frame. The dissertation consists of the introduction, 4 chapters, conclusions, list of literature. Introduction presents the novelty and topicality of the work, tasks and aims of the work are formulated. The first chapter is overview of voice quality evaluation methods, pros and cons of these methods are analyzed. PESQ algorithm and limits of its applicability are introduced in this chapter too. The lists of Lithuanian words for word intelligibility testing are created. Chapter two presents the method of signal construction that allows to extend PESQ applicability to short signals. This chapter introduces the concept of frame value. Distributions of frame values are calculated. Third chapter analyses distortions created by coding. It is shown that coding distortions... [to full text] / Disertacijoje nagrin jama koduoto balso kokybės vertinimo problematika. Pagrindinis dėmesys skiriamas balso kokybės tyrimams, kai perduodama koduota šneka ir prarandami balso paketai. Darbo tikslas yra patobulinti koduoto balso kokybės vertinimo algoritmus. Darbo uždaviniai yra šie: • sukurti matavimo priemonę trumpų balso signalo atkarpų kokybei vertinti; • apibrėžti koduoto balso segmentų vertės sampratą ir parinkti vertės metrikas; • išmatuoti bendrinės šnekos balso segmentų verčių skirstinius; • nustatyti skirtingų koderių sukuriamų iškraipymų ribas; • ištirti paplitusių koderių inertiškumą, nustatyti kiek laiko pastebima prarastų paketų įtaka sekantiems segmentams. Disertaciją sudaro įvadas, keturi tiriamieji skyriai ir bendrosios išvados. Įvade pristatomas darbo naujumas, aktualumas, aptariamas autoriaus indėlis, formuluojami darbo tikslai. Pirmas skyrius yra apžvalginis – analizuojami balso kokybės vertinimo metodai, jų privalumai ir trūkumai. Kaip savarankiška dalis čia pristatyti autoriaus sudaryti sąrašai lietuviškų žodžių, skirtų šnekos suprantamumo tyrimams. Antrame skyriuje parodoma, kaip galima išplėsti kokybės vertinimo PESQ (angl. Perceptual Evaluation of Speech Quality) algoritmo taikymo ribas. Čia įvedama koduoto balso paketo vertės sąvoka, nustatomi statistiniai paketų vertės skirstiniai. Trečiame skyriuje nagrinėjami specifiniai koduotos šnekos iškraipymai ir kodavimo parametrų įtaka... [toliau žr. visą tekstą]

FundaÃÃo Cearense de Apoio ao Desenvolvimento Cientifico e TecnolÃgico / Nos Ãltimos anos, observa-se o surgimento e a rÃpida disseminaÃÃo do serviÃo VoIP, integrando-se ao mercado atual junto à telefonia convencional e Ãs redes celulares. Por ser uma alternativa tecnolÃgica que contribui para minimizar ustos, assiste-se a uma preferÃncia crescente por fazer fegar a voz atravÃs das redes IP. O HSDPA, como sistema celular, permite a transmissÃo de dados em alta velocidade, aumenta a largura de banda da rede e abre novas possibilidades de serviÃos multimÃdia, como o VoIP que utiliza a transmissÃo em banda larga para telefones mÃveis. Exige-se, porÃm, um considerÃvel esforÃo de anÃlise deste serviÃo, pois o atraso inerente a esse sistema à um desafio para a garantia de qualidade de voz. Estes fatos justificam, conseqÃentemente, um esforÃo de anÃlise que se detenha sobre a qualidade de voz no VoIP sobre o HSDPA. Para avaliar a qualidade de voz, neste estudo aplica-se o mÃtodo MOS, que faz corresponder valores numÃricos a categorias como medidas de qualidade e inteligibilidade da voz transmitida, obtendo-se esses dados de forma objetiva e subjetiva. O processo de avaliaÃÃo dividiu-se em etapas de acordo com cada metodologia, seguindo recomendaÃÃes tÃcnicas e atravÃs de simulaÃÃes computacionais dinÃmicas. Na avaliaÃÃo objetiva, utilizou-se o algoritmo PESQ para obtenÃÃo do conceito MOS, enquanto que na avaliaÃÃo subjetiva, arquivos de voz com certo percentual de erro foram colocados em um endereÃo na Internet para escuta e atribuiÃÃo de nota MOS, baseada na percepÃÃo do usuÃrio ouvinte. Os resultados mostraram que os dois mÃtodos de avaliaÃÃo obtiveram conceitos de qualidade satisfatÃrios, que o QoS nas simulaÃÃes à estÃvel e positivo, que uma boa qualidade para os arquivos de voz e a provÃvel satisfaÃÃo dos usuÃrios do serviÃo VoIP sobre o sistema celular HSDPA à garantida para uma taxa de 2% de FER. Finalmente, mostra-se que a metodologia objetiva garante a obtenÃÃo de notas MOS aproximadas da subjetiva, evitando o Ãrduo trabalho de fazerem-se ouvir diversos arquivos de voz por uma quantidade significativa de usuÃrios para ser vÃlida estatisticamente. / In recent years, we can observe the development and fast dissemination of VoIP services, being integrated by the present market, beside conventional telephony and cellular networks. For being a technological alternative that contributes to minimize costs, we see an increasing preference for voice transmission through IP networks. HSDPA, as a cellular system, allows high speed data transmission,increases the network width of band and creates new possibilities for multimedia services, as VoIP that transmits in wideband to mobile telephones. The delay inherent to this system, however, is a challenge for the need to assure good quality of voice transmissions, demanding a considerable effort of analysis of this service.These facts justify a study that focus on the quality of voice in VoIP over HSDPA.To evaluate the voice quality, in this study, we applied MOS method that makes numerical values correspond to categories like quality and intelligibility of transmitted voice, getting these data through objective and subjective methodologies. The evaluation process was divided in fases according to the characteristics of each methodology and to technical recommendations, and was done through dynamic computational simulations. For objective evaluation process,algorithm PESQ was employed to obtain MOS concepts, whereas, for subjective evaluation, voice files with a percentage of error have been placed in Internet for listening and for the attribution of MOS concepts based in the perception of the listener. The results of this research show that both evaluation methods got satisfactory concepts of quality, that QoS is steady and positive in the simulations, that a good quality for the voice files and the probable satisfaction of the users of the VoIP service on cellular system HSDPA is guaranteed for 2% FER rate. Finally, it shows that MOS concepts produced by objective methodology were close enough to those given by subjective evaluation to dispense with the arduous work of making diverse voice files to be heard and subjectively evaluated by a statisticaly valid amount of users.

This degree project was made in collaboration with Nordicstation. The project task was to develop a testing application for a self-developed telephone survey service, which uses third party software. This third party software showed to be unreliable at higher loads. The purpose of the application is to analyse the speech quality of clients connected to the service. This report gives an introduction to the speech quality algorithms Perceptual Evaluation of Speech Quality (PESQ) and Single Sided Speech Quality Measure (3SQM). It also gives descriptions of the methods used to develop the application. The final chapters in this report are about the testing of the telephone service. The primary result of the testing was that the telephone service is unable to acceptably handle 80+ clients and recommendations to Nordicstation is to set a maximum of parallel connected clients to 80 or find an alternative to the third party software currently in use. / Detta examensarbete har gjorts i samarbete med Nordicstation. Projektets uppgift var att utveckla ett test program för at testa en egenutvecklad telefonundersökning-tjänst, baserad på tredjeparts mjukvara. Denna tredjeparts mjukvara visade sig vara opålitlig vid högre belastning. Syftet med programmet är att analysera samtals kvalitéten på de klienter som är anslutna till tjänsten. Denna rapport ger en introduktion till ljudkvalitetsalgoritmer så som Perceptual Evaluation of Speech Quality (PESQ) och Single Sided Speech Quality Measure (3SQM). Rapporten går även igenom de metoder som använts för att utvecklat programmet. De sista kapitlen i denna rapport är om själva testningen av telefonitjänsten. Det primära resultatet av testningen var att telefontjänsten inte kan hantera 80+ klienter acceptabelt och rekommendationer till Nordicstation är att sätta ett tak på maximalt parallellt anslutna klienter till 80 eller hitta ett alternativ till den tredjeparts mjukvara som nu används.

This project presents a mechanism for instrumentation of secure VoIP calls. The experiments were run under different network conditions and security systems. VoIP services such as Google Talk, Express Talk and Skype were under test. The project allowed analysis of the voice quality of the VoIP services based on the Mean Opinion Score (MOS) values generated by Perceptual valuation of Speech Quality (PESQ). The quality of the audio streams produced were subjected to end-to-end delay, jitter, packet loss and extra processing in the networking hardware and end devices due to Internetworking Layer security or Transport Layer security implementations. The MOS values were mapped to Perceptual Evaluation of Speech Quality for wideband (PESQ-WB) scores. From these PESQ-WB scores, the graphs of the mean of 10 runs and box and whisker plots for each parameter were drawn. Analysis on the graphs was performed in order to deduce the quality of each VoIP service. The E-model was used to predict the network readiness and Common vulnerability Scoring System (CVSS) was used to predict the network vulnerabilities. The project also provided the mechanism to measure the throughput for each test case. The overall performance of each VoIP service was determined by PESQ-WB scores, CVSS scores and the throughput. The experiment demonstrated the relationship among VoIP performance, VoIP security and VoIP service type. The experiment also suggested that, when compared to an unsecure IPIP tunnel, Internetworking Layer security like IPSec ESP or Transport Layer security like OpenVPN TLS would improve a VoIP security by reducing the vulnerabilities of the media part of the VoIP signal. Morever, adding a security layer has little impact on the VoIP voice quality.

<p>The acoustic echo problem arises whenever there is acoustic coupling between a loudspeaker and a microphone, such as in a teleconference system. This problem is traditionally solved by using an acoustic echo canceler (AEC), which models the echo path with adaptive filters. Long adaptive filters are necessary for satisfactory echo cancellation, which makes AEC highly computationally complex. Recently, a low-complexity echo suppression scheme was presented, the perceptual acoustic echo suppressor (PAES). Spectral modification is used to suppress the echoes, and the complexity is reduced by incorporating perceptual theories. However, under ideal conditions AEC performs better than PAES. </p><p>This thesis considers a hybrid system, which combines AEC and PAES. AEC is used to cancel low-frequency echo components, while PAES suppresses high-frequency echo components. The hybrid system is simulated and assessed, both through subjective listening tests and objective evaluations. The hybrid scheme is shown to have virtually the same perceived quality as a full-band AEC, while having a significantly lower complexity and a higher degree of robustness.</p>

The acoustic echo problem arises whenever there is acoustic coupling between a loudspeaker and a microphone, such as in a teleconference system. This problem is traditionally solved by using an acoustic echo canceler (AEC), which models the echo path with adaptive filters. Long adaptive filters are necessary for satisfactory echo cancellation, which makes AEC highly computationally complex. Recently, a low-complexity echo suppression scheme was presented, the perceptual acoustic echo suppressor (PAES). Spectral modification is used to suppress the echoes, and the complexity is reduced by incorporating perceptual theories. However, under ideal conditions AEC performs better than PAES. This thesis considers a hybrid system, which combines AEC and PAES. AEC is used to cancel low-frequency echo components, while PAES suppresses high-frequency echo components. The hybrid system is simulated and assessed, both through subjective listening tests and objective evaluations. The hybrid scheme is shown to have virtually the same perceived quality as a full-band AEC, while having a significantly lower complexity and a higher degree of robustness.

Popularity of speech communication is rapidly increasing in various contexts such as conferencing systems, mobile/fixed electronic devices and laptops thus leading to a heightened demand for new services and improved speech quality. Dictaphones used for dictations usually have one microphone. Single microphone does not give enough degree of freedom to allow estimation of location of the source. Microphone array makes use of multiple microphones for spatial filtering suppressing the background noise. This report aims for speech enhancement utilizing the benefits inherited with microphone arrays to find direction of desired speaker and focus the listening beam in that direction. A comparison is made between Generalized Cross Correlation (GCC) methods for locating the source in real office environment. Beamforming is implemented to make the microphone array listen in the desired direction thus reducing the interference from other sources. Minimum Variance Distortion-less Response (MVDR) approach is shown to give better results compared to more simplistic techniques. Perceptual based Eigen filter incorporating human hearing models in subspace incorporated in the suppressor eliminates the residual noise. Objective system performance is evaluated by estimating Signal-to-Noise-Ratio improvement (SNRI), segmental SNR, signal degradation and noise suppression. Perpetual Evaluation of Speech Quality (PESQ) gives Mean Opinion Score for subjective evaluation. / asim_zolo@yahoo.com, akbarali45@gmail.com

In communication system environment speech signal is corrupted due to presence of additive acoustic noise, so with this distortion the effective communication is degraded in terms of the quality and intelligibility of speech. Now present research is going how effectively acoustic noise can be eliminated without affecting the original speech quality, this tends to be our challenging in this current research thesis work. Here this work proposes multi-tiered detection method that is based on time-frequency analysis (i.e. filter banks concept) of the noisy speech signals, by using standard speech enhancement method based on the proven spectral subtraction, for single channel speech data and for a wide range of noise types at various noise levels. There were various variants have been introduced to standard spectral subtraction proposed by S.F.Boll. In this thesis we designed and implemented a novel approach of Spectral Subtraction based on Minimum Statistics [MinSSS]. This means that the power spectrum of the non-stationary noise signal is estimated by finding the minimum values of a smoothed power spectrum of the noisy speech signal and thus circumvents the speech activity detection problem. This approach is also capable of dealing with non-stationary noise signals. In order to analyze the system in time frequency domain, we have implemented two different filter bank approaches such as Weighted OverLap Added (WOLA) and Fast Fourier Transform Modulated (FFTMod). The proposed systems were implemented and evaluated offline using simulation tool Matlab and then validated their performances based on the objective quality measures such as Signal to Noise Ratio Improvement (SNRI) and Perceptual Evaluation Speech Quality (PESQ) measure. The systems were tested with a pure speech combination of male and female sampled at 8 kHz, these signals were corrupted with various kinds of noises at different noise power levels. The MinSSS algorithm implemented using FFTMod filter bank approach outperforms when compared the WOLA filter bank approach.

This thesis deals with description of human hearing, audio and speech codecs, description of objective measure of quality and practical comparison of codecs. Chapter about audio codecs consists of description of lossless codec FLAC and lossy codecs MP3 and Ogg Vorbis. In chapter about speech codecs is description of linear predictive coding and G.729 and OPUS codecs. Evaluation of quality consists of description of segmental signal-to- noise ratio and perceptual evaluation of quality – WSS and PESQ. Last chapter deals with description od practical part of this thesis, that is comparison of memory and time consumption of audio codecs and perceptual evaluation of speech codecs quality.