Search

Both technological components of PSTN and VoIP speech transmission as well as aspects of the perception of transmitted speech are systematically described in this chapter. It is explained how quality is formed on the basis of perceived dimensions, how quality and quality dimensions can subjectively be measured, and how quality can be modeled on the basis of dimensions. State-of-the-art signal-based and parametric quality models are presented and the advantages of a new class of models with diagnostic capabilities are stressed. Finally, the research topics of this book are formulated: The identification of relevant quality dimensions, the development of an efficient subjective test method for dimension assessment, and the development of a parametric dimension-based model for speech quality.

Automatic Speech Recognition (ASR) systems are increasing in usage for voice centric applications in mobile handheld and Voice over Internet Protocol (VoIP) based devices. The necessity is also increasing to find out the ASR performance under different network impediments when the recognition is performed in the remote servers, in real sense. Among the major impediments, speech coding is the one, which affects the ASR performance greatly, when using it with different sampling rates and bit rates in the practical systems. The speech codecs which use different algorithms for generating different bit rates will affect the speech sounds, i.e. vowels and consonants, differently, and cause the critical sounds in the words to be changed and in-turn affects the overall word recognition performance of the ASR systems. In this paper, the influence of the sampling rate and bit rate changes with different narrowband and wideband codecs on the speech sounds is analyzed. Investigation is carried out to see how the speech sounds are changing while using different codecs operating at different bit rates.

This paper proposes a quality-aware loss-robust scalable speech streaming (QLSSS) method to improve the perceived speech quality (PSQ) of a scalable wideband speech streaming (SWSS) system over IP networks. To this end, the proposed method estimates the PSQ and the packet loss rate (PLR) from the received speech data. Subsequently, it decides the amount of redundant speech data (RSD) that a speech decoder can use to reconstruct lost speech signals for high PLRs. According to this decision, the proposed method optimizes a scalable speech coding mode for current speech data (CSD) and RSD bitstreams in order to prevent speech quality from being degraded under the estimated packet loss condition and maintain the transmission bandwidth. The effectiveness of the proposed method is then demonstrated using the ITU-T Recommendations G.729.1 and P.563 as a scalable wideband speech codec and a PSQ estimator, respectively. It is shown from the experiments that an SWSS system employing the proposed QLSSS method significantly improves speech quality under packet loss conditions.

Since the early 1980s, advances in speech coding technologies have enabled speech coders to achieve bit-rate reductions of a factor of 4 to 8 while maintaining roughly the same high speech quality. One of the most important driving forces behind this feat is the so-called analysis-by-synthesis paradigm for coding the excitation signal of predictive speech coders. In this chapter, we give an overview of many variations of the analysis-by-synthesis excitation coding paradigm as exemplified by various speech coding standards around the world. We describe the variations of the same basic theme in the context of different coder structures where these techniques are employed. We also attempt to show the relationship between them in the form of a family tree. The goal of this chapter is to give the readers a big-picture understanding of the dominant types of analysis-by-synthesis excitation coding techniques for predictive speech coding.

Low-bit-rate speech coding, at rates below 4 kb/s, is needed for both communication and voice storage applications. At such low rates, full encoding of the speech waveform is not possible; therefore, low-rate coders rely instead on parametric models to represent only the most perceptually relevant aspects of speech. While there are a number of different approaches for this modeling, all can be related to the basic linear model of speech production, where an excitation signal drives a vocal-tract filter.

Speech intelligibility is the most important parameter in evaluation of speech quality. In the contribution, a new objective intelligibility assessment of general speech processing algorithms is proposed. It is based on automatic recognition methods developed for discrete and fluent speech processing. The idea is illustrated on two case studies: a) comparison of listening evaluation of Czech rhyme tests with automatic discrete speech recognition and b) automatic continuous speech recognition of general topic Czech texts read by professional and nonprofessional speakers vs. the same texts generated by several Czech Text-to-Speech systems. The aim of the proposed approach is fast and objective intelligibility assessment of Czech Text-to-Speech systems, which include male and female voices and a voice conversion module.

The emerging applications of wireless speech communication are demanding increasing levels of performance in noise adverse environments together with the design of high response rate speech processing systems. This is a serious obstacle to meet the demands of modern applications and therefore these systems often needs a noise reduction algorithm working in combination with a precise voice activity detector (VAD). This paper presents a new voice activity detector (VAD) for improving speech detection robustness in noisy environments and the performance of speech recognition systems. The algorithm defines an optimum likelihood ratio test (LRT) involving Multiple and correlated Observations (MCO). An analysis of the methodology for N = {2, 3} shows the robustness of the proposed approach by means of a clear reduction of the classification error as the number of observations is increased. The algorithm is also compared to different VAD methods including the G.729, AMR and AFE standards, as well as recently reported algorithms showing a sustained advantage in speech/non-speech detection accuracy and speech recognition performance.

The emerging applications of wireless speech communication are demanding increasing levels of performance in noise adverse environments together with the design of high response rate speech processing systems. This is a serious obstacle to meet the demands of modern applications and therefore these systems often needs a noise reduction algorithm working in combination with a precise voice activity detector (VAD). This paper presents a new voice activity detector (VAD) for improving speech detection robustness in noisy environments and the performance of speech recognition systems. The algorithm defines an optimum likelihood ratio test (LRT) involving Multiple and correlated Observations (MO) and assuming a jointly Gaussian probability density function (jGpdf). An analysis of the methodology for N = {2, 3} shows the robustness of the proposed approach by means of a clear reduction of the classification error as the number of observations is increased. The algorithm is also compared to different VAD methods including the G.729, AMR and AFE standards, as well as recently reported algorithms showing a sustained advantage in speech/non-speech detection accuracy and speech recognition performance.

The concepts and experiments presented are focused on modifications of an existing parametric speech coding algorithm (CELP) introduced in order to improve subjective speech quality in telephone connections. The perceptual coding to bit rate limiting was added and algorithms qualifying speech components to the categories of "voiced", "unvoiced", "transients" using rough sets were studied. The speech signal quality achieved with the proposed hybrid codec was compared to the quality offered by some standard speech codecs.

Speech quality, as perceived by the users of Voice over Internet Protocol (VoIP) telephony, is critically important to the uptake of this service. VoIP quality can be degraded by network layer problems (delay, jitter, packet loss). This paper presents a method for real-time, non-intrusive speech quality estimation for VoIP that emulates the subjective listening quality measures based on Mean Opinion Scores (MOS). MOS provide the numerical indication of perceived quality of speech. We employ a Genetic Programming based symbolic regression approach to derive a speech quality estimation model. Our results compare favorably with the International Telecommunications Union-Telecommunication Standardization (ITU-T) PESQ algorithm which is the most widely accepted standard for speech quality estimation. Moreover, our model is suitable for real-time speech quality estimation of VoIP while PESQ is not. The performance of the proposed model was also compared to the new ITU-T recommendation P.563 for non-intrusive speech quality estimation and an improved performance was observed.

Voice packets with guaranteed QoS (Quality of Service) on the VoIP system are responsible for digitizing, encoding, decoding, and playing out the speech signal. The important point is based on the factor that different parts of speech over IP networks have different perceptual importance and each part of speech does not contribute equally to the overall voice quality. In this paper, we propose new additive noise reduction algorithms to improve voice over IP networks and present performance evaluation of perceptual speech signal through IP networks in the additive noise environment during realtime phone-call service. The proposed noise reduction algorithm is applied to pre-processing method before speech coding and to post-processing method after speech decoding based on single microphone VoIP system. For noise reduction, this paper proposes a Wiener filter optimized to the estimated SNR of noisy speech for speech enhancement. Various noisy conditions including white Gaussian, office, babble, and car noises are considered with G.711 codec. Also, we provide critical message report procedures and management schemes to guarantee QoS over IP networks. Finally, as following the experimental results, the proposed algorithm and method has been prove for improving speech quality.

Tracheoesophageal (TE) speech is a possibility to restore the ability to speak after laryngectomy, i.e. the removal of the larynx. TE speech often shows low audibility and intelligibility which also makes it a challenge to automatic speech recognition. We improved the recognition results by adapting a speech recognizer trained on normal, non-pathologic voices to single TE speakers by unsupervised HMM interpolation.

In-Car speech recognition will be pervasive over the coming years. The goal of speech enhancement is to increase the quality and intelligibility of speech in a noisy environment. The focus of the present research is to evaluate the effect of speech enhancement on the intelligibility of spoken language in a moving vehicle. Here, an ECoS network is used as a model to evaluate the intelligibility. A baseline performance was established using clean speech data. This data was then mixed with various types of in-vehicle noise at several signal-to-noise ratios. Speech enhancement techniques were applied to the noisy speech data. The performance of the ECoS model was evaluated when the noisy and enhanced speech was presented. Several factors were found to affect the recognition rate, including noise type and noise volume.

A simulation-based methodology is developed for analyzing the subjective quality of voice-over-IP calls as a function of network QoS parameters and choices in implementation and configuration. The proposed method combines the use of existing objective voice-quality measurement algorithms, such as the ITU-T P.861 PSQM, and artificial voice reference signals, such as the ITU-T P.50, with the discrete-event simulation of a network QoS model. A significant advantage of the method is that it does not involve the use of human subjects in evaluating subjective voice quality levels. This enables one to entirely automate the process of quantifying call quality as a function of network QoS and implementation choices such as packet size and codec type. Such automation enables one to realize significant time and cost savings in obtaining experimental results. A tool implementation is described that includes basic network packet loss and delay jitter models. Example numerical results are presented for the G.711 codec. The extension of the method to the subjective quality evaluation of audio, video, and multimedia signals is also described. The paper also introduces the concepts of 'subjective teletraffic engineering' and 'subjective equivalent bandwidths.'