Perceptual Speech Quality Research Articles

Practical Consistency Between Bit-Error and Block-Error Performance Metrics up to Application Layer

Although the 3G (UMTS) systems physical-layer performance is still described by both bit-error-rate and block-error-rate, the 4G (LTE) uses exclusively the latter that is at higher protocol layers commonly referred to as frame-error-rate or packet-error-rate of downlink and uplink input data streams. However, the block-error-rate estimation scheme heavily depends on the success rate of negative acknowledgements transport via return channel, and so does not provide trustworthy low values (e.g. around $$10^{-5})$$10-5). The remaining errors that are not corrected by the physical layer, determine the so-called residual channel and need to be corrected by higher protocols to provide almost errorless application environment. Error rate estimation is therefore important not only in research, development and manufacturing, but also in network operator's environment. However, relationship between the bit-error and the block-error oriented performance metrics is not always a simple one, even at physical layer alone, as well as through higher-protocol layers. In this paper, we benchmark in practice both types of performance metrics for consistency, specifically on an operational microwave link residual channel, bottom upwards to the application-layer, end-to-end perceptual speech quality measurements.

Objective speech intelligibility prediction in sensorineural hearing loss using acoustic simulations and perceptual speech quality measures

A novel approach to objectively predict speech intelligibility in sensorineural hearing loss using acoustic simulations of impaired perception and objective measures of perceptual speech quality (PESQ) is proposed and validated. Acoustic simulations of impaired perception with different types and degrees of hearing loss were obtained degrading of the original speech waveforms by spectral smearing, expansive nonlinearity, and level scaling. The CUNY NST syllables were used as test material. PESQ was used to measure perceptual quality of the acoustic simulations thus obtained. Finally, PESQ scores were transformed into predicted intelligibility scores using a logistic function. Validation of the proposed objective method was performed by comparing predicted intelligibility scores with subjective measures of intelligibility of the degraded speech in a group of ten subjects. Predictive intelligibility scores showed good correlation (R² = 0.7) with subjective intelligibility scores and a low error in the prediction (RMSE = 0.14). The proposed approach could be a valuable aid in real clinical applications where it is needed to measure speech intelligibility and might be of some help in avoiding time-consuming experimental measurements. In particular, this method might be valuable in the characterization of the sensitivity of new speech tests for screening and diagnosing of hearing loss, or in the assessment of the performance of novel algorithms of speech enhancement for a target hearing impairment.

Are you hearing voices in the high frequencies of human speech and voice?

The human voice produces acoustic energy at frequencies above 6 kHz. Energy in this high-frequency region has long been known to affect perception of speech and voice quality, but also provides non-qualitative information about a speech signal. This presentation will demonstrate how much useful information can be gleaned from the high frequencies with a report on studies where listeners were presented with only high-frequency energy extracted from speech and singing. Come to test your own abilities and decide if you can hear strange voices or just chirps and whistles in the high frequencies of human speech and voice.

Subband DCT and EMD Based Hybrid Soft Thresholding for Speech Enhancement

This paper presents a two-stage soft thresholding algorithm based on discrete cosine transform (DCT) and empirical mode decomposition (EMD). In the first stage, noisy speech is decomposed into eight frequency bands and a specific noise variance is calculated for each one. Based on this variance, each band is denoised using soft thresholding in DCT domain. The remaining noise is eliminated in the second stage through a time domain soft thresholding strategy adapted to the intrinsic mode functions (IMFs) derived by applying EMD on the signal obtained from the first stage processing. Significantly better SNR improvement and perceptual speech quality results for different noise types prove the superiority of the proposed algorithm over recently reported techniques.

Speech Intelligibility Prediction Intended for State-of-the-Art Noise Estimation Algorithms

Noise estimation is critical factor of any speech enhancement system. In presence of additive non-stationary background noise, it is difficult to understand speech for normal hearing particularly for hearing impaired person. The background interfering noise reduces the intelligibility and perceptual quality of speech. Speech enhancement with various noise estimation techniques attempts to minimize the interfering components and enhance the intelligibility and perceptual aspects of damaged speech. This study addresses the selection of right noise estimation algorithm in speech enhancement system for intelligent hearing. A noisy environment of airport is considered. The clean speech is corrupted by noisy environment for different noise levels ranging from 0 to 15 dB. Six diverse noise estimation algorithms are selected to estimate the noise including Minimum Controlled Recursive Average (MCRA), MCRA-2, improved MCRA, Martin minimum tracking, continuous spectral minimum tracking, and weighted spectral average. Spectral subtraction algorithm is used for enhancing the noisy speech. The intelligibility of enhanced speech is assessed by the fractional Articulation Index (fAI) and SNR_LOSS.

The effects of noise vocoding on speech quality perception

Speech perception depends on access to spectral and temporal acoustic cues. Temporal cues include slowly varying amplitude changes (i.e. temporal envelope, TE) and quickly varying amplitude changes associated with the center frequency of the auditory filter (i.e. temporal fine structure, TFS). This study quantifies the effects of TFS randomization through noise vocoding on the perception of speech quality by parametrically varying the amount of original TFS available above 1500Hz. The two research aims were: 1) to establish the role of TFS in quality perception, and 2) to determine if the role of TFS in quality perception differs between subjects with normal hearing and subjects with sensorineural hearing loss. Ratings were obtained from 20 subjects (10 with normal hearing and 10 with hearing loss) using an 11-point quality scale. Stimuli were processed in three different ways: 1) A 32-channel noise-excited vocoder with random envelope fluctuations in the noise carrier, 2) a 32-channel noise-excited vocoder with the noise-carrier envelope smoothed, and 3) removal of high-frequency bands. Stimuli were presented in quiet and in babble noise at 18dB and 12dB signal-to-noise ratios. TFS randomization had a measurable detrimental effect on quality ratings for speech in quiet and a smaller effect for speech in background babble. Subjects with normal hearing and subjects with sensorineural hearing loss provided similar quality ratings for noise-vocoded speech.

Effects of various electrode configurations on music perception, intonation and speaker gender identification

Advances in speech coding strategies and electrode array designs for cochlear implants (CIs) predominantly aim at improving speech perception. Current efforts are also directed at transmitting appropriate cues of the fundamental frequency (F0) to the auditory nerve with respect to speech quality, prosody, and music perception. The aim of this study was to examine the effects of various electrode configurations and coding strategies on speech intonation identification, speaker gender identification, and music quality rating. In six MED-EL CI users electrodes were selectively deactivated in order to simulate different insertion depths and inter-electrode distances when using the high definition continuous interleaved sampling (HDCIS) and fine structure processing (FSP) speech coding strategies. Identification of intonation and speaker gender was determined and music quality rating was assessed. For intonation identification HDCIS was robust against the different electrode configurations, whereas fine structure processing showed significantly worse results when a short electrode depth was simulated. In contrast, speaker gender recognition was not affected by electrode configuration or speech coding strategy. Music quality rating was sensitive to electrode configuration. In conclusion, the three experiments revealed different outcomes, even though they all addressed the reception of F0 cues. Rapid changes in F0, as seen with intonation, were the most sensitive to electrode configurations and coding strategies. In contrast, electrode configurations and coding strategies did not show large effects when F0 information was available over a longer time period, as seen with speaker gender. Music quality relies on additional spectral cues other than F0, and was poorest when a shallow insertion was simulated.

Regularization for Partial Multichannel Equalization for Speech Dereverberation

Acoustic multichannel equalization techniques such as the multiple-input/output inverse theorem (MINT), which aim to equalize the room impulse responses (RIRs) between the source and the microphone array, are known to be highly sensitive to RIR estimation errors. To increase robustness, it has been proposed to incorporate regularization in order to decrease the energy of the equalization filters. In addition, more robust partial multichannel equalization techniques such as relaxed multichannel least-squares (RMCLS) and channel shortening (CS) have recently been proposed. In this paper, we propose a partial multichannel equalization technique based on MINT (P-MINT) which aims to shorten the RIR. Furthermore, we investigate the effectiveness of incorporating regularization to further increase the robustness of P-MINT and the aforementioned partial multichannel equalization techniques, i.e., RMCLS and CS. In addition, we introduce an automatic non-intrusive procedure for determining the regularization parameter based on the L-curve. Simulation results using measured RIRs show that incorporating regularization in P-MINT yields a significant performance improvement in the presence of RIR estimation errors, whereas a smaller performance improvement is observed when incorporating regularization in RMCLS and CS. Furthermore, it is shown that the intrusively regularized P-MINT technique outperforms all other investigated intrusively regularized multichannel equalization techniques in terms of perceptual speech quality (PESQ). Finally, it is shown that the automatic non-intrusive regularization parameter in regularized P-MINT leads to a very similar performance as the intrusively determined optimal regularization parameter, making regularized P-MINT a robust, perceptually advantageous, and practically applicable multichannel equalization technique for speech dereverberation.

Single-trial analysis of the neural correlates of speech quality perception

Objective. Assessing speech quality perception is a challenge typically addressed in behavioral and opinion-seeking experiments. Only recently, neuroimaging methods were introduced, which were used to study the neural processing of quality at group level. However, our electroencephalography (EEG) studies show that the neural correlates of quality perception are highly individual. Therefore, it became necessary to establish dedicated machine learning methods for decoding subject-specific effects. Approach. The effectiveness of our methods is shown by the data of an EEG study that investigates how the quality of spoken vowels is processed neurally. Participants were asked to indicate whether they had perceived a degradation of quality (signal-correlated noise) in vowels, presented in an oddball paradigm. Main results. We find that the P3 amplitude is attenuated with increasing noise. Single-trial analysis allows one to show that this is partly due to an increasing jitter of the P3 component. A novel classification approach helps to detect trials with presumably non-conscious processing at the threshold of perception. We show that this approach uncovers a non-trivial confounder between neural hits and neural misses. Significance. The combined use of EEG signals and machine learning methods results in a significant ‘neural’ gain in sensitivity (in processing quality loss) when compared to standard behavioral evaluation; averaged over 11 subjects, this amounts to a relative improvement in sensitivity of 35%.

Collaborative Blind Source Separation Using Location Informed Spatial Microphones

This letter presents a new Collaborative Blind Source Separation (CBSS) technique that uses a pair of location informed coincident microphone arrays to jointly separate simultaneous speech sources based on time-frequency source localization estimates from each microphone recording. While existing BSS approaches are based on localization estimates of sparse time-frequency components, the proposed approach can also recover non-sparse (overlapping) time-frequency components. The proposed method has been evaluated using up to three simultaneous speech sources under both anechoic and reverberant conditions. Results from objective and subjective measures of the perceptual quality of the separated speech show that the proposed approach significantly outperforms existing BSS approaches.

Speech Quality Assessment of VoIP: G.711 VS G.722 Based on Interview Tests with Thai Users

This paper presents the comparison between two codecs, G.711 and G.722 at 64 kbps, referring to speech quality perception using a subjective method called interview tests. These subjective tests have been conducted with 201 subjects, who are Thai native speakers that use Thai which is a tonal language, for accuracy and reliability of results. The results from testing with both codecs are almost the same; the scores are 4.17 for G.722 and 4.14 for G.711. After analyzing the results, it has been confirmed that G.722 does not provide better speech quality than G.711 to the Thai subjects significantly, which is consistent with previous information. However, these results could be used as the benchmark of G.711 and G.722 for speech quality assessment within Thai environments.

An Investigation on Speech Perception Under Effects of Coarticulation

In this paper, we investigated the speech perception under effects of coarticulation within syllables, including estimating the nuclei and coarticulated transition intervals of phonemes, and investigating the perception of nuclei and coarticulated transition intervals of phonemes within syllables. To complete these objectives, we firstly proposed the folded spectral transition measure to estimate the boundary points between the nuclei interval and coarticulated transition intervals. Then, we conducted experiments to investigate the syllable identification and syllable quality evaluation under effects of coarticulation. The experimental results show that human can identify the syllables with the retaining intervals equal or wider the coarticulated transition intervals. Besides, the experimental results show that the nuclei intervals are still important for speech quality perception; the quality of truncated syllables retaining the two phoneme nucleus and coarticulated transition interval between the two phonemes within syllable is equivalent with the quality of original two-phoneme syllables.

Adaptive Redundant Speech Transmission over Wireless Multimedia Sensor Networks Based on Estimation of Perceived Speech Quality

An adaptive redundant speech transmission (ARST) approach to improve the perceived speech quality (PSQ) of speech streaming applications over wireless multimedia sensor networks (WMSNs) is proposed in this paper. The proposed approach estimates the PSQ as well as the packet loss rate (PLR) from the received speech data. Subsequently, it decides whether the transmission of redundant speech data (RSD) is required in order to assist a speech decoder to reconstruct lost speech signals for high PLRs. According to the decision, the proposed ARST approach controls the RSD transmission, then it optimizes the bitrate of speech coding to encode the current speech data (CSD) and RSD bitstream in order to maintain the speech quality under packet loss conditions. The effectiveness of the proposed ARST approach is then demonstrated using the adaptive multirate-narrowband (AMR-NB) speech codec and ITU-T Recommendation P.563 as a scalable speech codec and the PSQ estimation, respectively. It is shown from the experiments that a speech streaming application employing the proposed ARST approach significantly improves speech quality under packet loss conditions in WMSNs.

New Results on Single-Channel Speech Separation Using Sinusoidal Modeling

We present new results on single-channel speech separation and suggest a new separation approach to improve the speech quality of separated signals from an observed mixture. The key idea is to derive a mixture estimator based on sinusoidal parameters. The proposed estimator is aimed at finding sinusoidal parameters in the form of codevectors from vector quantization (VQ) codebooks pre-trained for speakers that, when combined, best fit the observed mixed signal. The selected codevectors are then used to reconstruct the recovered signals for the speakers in the mixture. Compared to the log-max mixture estimator used in binary masks and the Wiener filtering approach, it is observed that the proposed method achieves an acceptable perceptual speech quality with less cross-talk at different signal-to-signal ratios. Moreover, the method is independent of pitch estimates and reduces the computational complexity of the separation by replacing the short-time Fourier transform (STFT) feature vectors of high dimensionality with sinusoidal feature vectors. We report separation results for the proposed method and compare them with respect to other benchmark methods. The improvements made by applying the proposed method over other methods are confirmed by employing perceptual evaluation of speech quality (PESQ) as an objective measure and a MUSHRA listening test as a subjective evaluation for both speaker-dependent and gender-dependent scenarios.

Iterative Kalman Versus Rao-blackwellized Particle Filter in Speech Enhancement

In this paper, the performance of two single channel speech enhancement filters: iterative Kalman and Rao-blackwellized particle filters, are evaluated. Speech intelligibility is deliberately deteriorated by adding white, colored and industrial noises. The quality of the processed speech is assessed by the Perceptual Quality of Speech (PESQ) index. Particular attention is also devoted to the reduction of the implementation cost, by assigning different model order and particle number to the voiced and unvoiced frames. The simulation results show mixed outcome. In some cases IKF and in some others RBPF gives better PESQ scores. The quality of each actually depends on the type of noise and input SNR. From viewpoint of computational costs, of course, IKF is more efficient than RBPF.

Effects of Noise, Nonlinear Processing, and Linear Filtering on Perceived Speech Quality

The purpose of this study was to measure subjective quality ratings in listeners with normal hearing and listeners with hearing loss for speech subjected to a wide range of processing conditions that are representative of real hearing aids. Speech quality was assessed using a rating scale in a group of 14 listeners with normal hearing and 15 listeners with mild to moderately severe sensorineural hearing loss. Controlled simulations of hearing aid processing were used to process speech that included speech subjected to (1) noise and nonlinear processing, (2) linear filtering, and (3) combinations of noise, nonlinear processing, and linear filtering. The 32 conditions of noise and nonlinear processing included stationary speech-shaped nose, multitalker babble, peak clipping, quantization noise, spectral subtraction, and dynamic range compression (in quiet, with babble, and with spectral subtraction). The 32 linear filtering conditions included high-pass filtering, low-pass filtering, band-pass filtering, positive and negative spectral tilt, and resonance peaks. Subsets of these conditions were used for the 36 conditions that combined noise and nonlinear processing with linear processing. Both listeners with normal hearing and listeners with hearing loss gave consistent (reliable) ratings. In both listener groups, sound quality was significantly affected by the noise, nonlinear processing, and linear filtering conditions. Compared with the listeners with normal hearing, the listeners with hearing loss showed significantly lower ratings of sound quality in nearly all of the processing conditions. For the conditions included in the current hearing aid simulation, noise and nonlinear conditions had a greater effect on quality judgments than did the linear filtering conditions. The data reported here provide a comprehensive dataset of speech quality ratings for simulated hearing aid processing conditions. The results indicate that quality ratings by listeners with hearing loss are significantly lower than quality ratings by listeners with normal hearing. In addition, quality ratings by listeners with hearing loss are impacted by signal processing at least as much as, and often more than, the quality ratings by listeners with normal hearing. Finally, quality ratings for speech processed with a simulated hearing aid are impacted more by noise and nonlinear signal processing than by linear filtering.

Improving Perceptual Quality of Speech in a Noisy Environment by Enhancing Temporal Envelope and Pitch

This letter proposes a new method of improving perceptual quality of speech in a noisy environment. Based on the perception theory that the temporal envelope plays a major role in assessing intelligibility, the proposed method uses a novel operation that enhances the fluctuation of the band-wise temporal envelope. It also contains pitch enhancement in frequency domain to further improve the perceptual quality. A new subjective evaluation scheme employing binaural listening is proposed in order to measure perceptual quality more reliably. The subjective evaluation shows that the proposed method improves perceptual quality of speech in various noisy environments.

SPEECH ENHANCEMENT FOR ROBUST SPEECH RECOGNITION IN MOTORCYCLE ENVIRONMENT

In the present work, we investigate the performance of a number of traditional and recent speech enhancement algorithms in the adverse non-stationary conditions, which are distinctive for motorcycles on the move. The performance of these algorithms is ranked in terms of the improvement they contribute to the speech recognition accuracy, when compared to the baseline performance, i.e. without speech enhancement. The experiments on the MoveOn motorcycle speech and noise database indicated that there is no equivalence between the ranking of algorithms based on the human perception of speech quality and the speech recognition performance. The Multi-band spectral subtraction method was observed to lead to the highest speech recognition performance.

A Cross Layer Perceptual Speech Quality Based Wireless VoIP Service

This paper proposes a cross layer wireless VoIP service which integrates an Adaptive QoS Playout (AQP) algorithm, E-model, Stream Control Transmission Protocol (SCTP), IEEE 802.21 Media Independent Handover (MIH) middleware and two user motion detection services. The proposed AQP algorithm integrates the effect of playout control and lost packet retransmission based on the E-model. Besides, by using the partial reliable transmission service from SCTP and the handoff notification from MIH services in a cross layer manner, AQP can reduce the lateness loss rate and improve speech quality under high frame error rates. In the simulations, the performance of AQP is compared with a fixed playout algorithm and four adaptive playout strategies. The simulation results show that the lateness loss rate of AQP is 2% lower than that of existing playout algorithms and the R-factor is 16% higher than the compared algorithms when a network has 50ms wired propagation delay and 2.5% frame error rate.

Evolutionary speech quality estimation in VoIP

Estimating the quality of Voice over Internet Protocol (VoIP) as perceived by humans is considered a formidable task. This is partly due to the relatively large number of variables that are involved as determinants of quality. Moreover, discerning the significance of one variable over the other is difficult. In this paper a novel approach based on genetic programming (GP) is presented. It maps the effect of network traffic parameters on listeners’ perception of speech quality. The ITU-T Recommendation P.862 (PESQ) algorithm is used as a reference model in this research. The GP discovered models that provide effective VoIP quality estimation are highly correlated to ITU-T Recommendation P.862 (PESQ). They also outperform the ITU-T Recommendation P.563 in estimating the effect that packet loss has on speech quality. The GP discovered models prove suited to real-time and in vivo evaluation of VoIP calls. Additionally, they are deployable on a wide variety of hardware platforms.