Search

Automatically predicting the outcome of subjective listening tests is a challenging task. Ratings may vary from person to person even if preferences are consistent across listeners. While previous work has focused on predicting listeners' ratings (mean opinion scores) of individual stimuli, we focus on the simpler task of predicting subjective preference given two speech stimuli for the same text. We propose a model based on anti-symmetric twin neural networks, trained on pairs of waveforms and their corresponding preference scores. We explore both attention and recurrent neural nets to account for the fact that stimuli in a pair are not time aligned. To obtain a large training set we convert listeners' ratings from MUSHRA tests to values that reflect how often one stimulus in the pair was rated higher than the other. Specifically, we evaluate performance on data obtained from twelve MUSHRA evaluations conducted over five years, containing different TTS systems, built from data of different speakers. Our results compare favourably to a state-of-the-art model trained to predict MOS scores.

Objectives: Many cochlear implant (CI) recipients use a hearing aid (HA) on the non-implanted ear. Current fitting rules of HAs are optimized on speech rather than music as input. However, a potential improvement might be the transmission of low-frequency acoustic sounds particularly targeting the requirements [zum vollständigen Text gelangen Sie über die oben angegebene URL]

Non-interactive and linear experienceslike cinema film offer high quality surround sound audio to enhance immersion, however, the perspective is usually fixed to the recording microphone position. With the rise of virtual reality, there is a demand for recording and recreating real-world experiences that allow users to move throughout the reproduction. Sound field translation achieves this by building an equivalent environment of virtual sources to recreate the recording spatially. However, the technique remains to restrict the maximum distance a user can translate away from the recording microphone's perspective due to the discrete sampling by commercial higher order microphones only being capable of recording an acoustic sweet-spot. In this paper, we propose a method for binaurally reproducing a microphone recording in a virtual application that allows the user to freely translate their body further beyond the recording position. The method incorporates a mixture of near-field and far-field sources in a sparsely expanded virtual environment to maintain a perceptually accurate reproduction. We perceptually validate the method through a Multiple Stimulus with Hidden Reference and Anchor (MUSHRA) experiment. Compared to the planewave benchmark, the proposed method offers both improved source localizability and robustness to spectral distortions at translated listening positions. A cross-examination with numerical simulations demonstrated that the sparse expansion relaxes the inherent sweet-spot constraint, leading to the improved localizability for sparse environments. Additionally, the proposed method is seen to better reproduce the intensity and binaural room impulse response spectra of near-field environments, further supporting the perceptual results.

Text-to-speech (TTS) synthesis systems have been widely used in general-purpose applications based on the generation of speech. Nonetheless, there are some domains, such as storytelling or voice output aid devices, which may also require singing. To enable a corpus-based TTS system to sing, a supplementary singing database should be recorded. This solution, however, might be too costly for eventual singing needs, or even unfeasible if the original speaker is unavailable or unable to sing properly. This work introduces a unit selection-based text-to-speech-and-singing (US-TTS&S) synthesis framework, which integrates speech-to-singing (STS) conversion to enable the generation of both speech and singing from an input text and a score, respectively, using the same neutral speech corpus. The viability of the proposal is evaluated considering three vocal ranges and two tempos on a proof-of-concept implementation using a 2.6-h Spanish neutral speech corpus. The experiments show that challenging STS transformation factors are required to sing beyond the corpus vocal range and/or with notes longer than 150 ms. While score-driven US configurations allow the reduction of pitch-scale factors, time-scale factors are not reduced due to the short length of the spoken vowels. Moreover, in the MUSHRA test, text-driven and score-driven US configurations obtain similar naturalness rates of around 40 for all the analysed scenarios. Although these naturalness scores are far from those of vocaloid, the singing scores of around 60 which were obtained validate that the framework could reasonably address eventual singing needs.

Objectives: Cochlear implants (CIs) restore speech perception in quiet but they also eliminate or distort many acoustic cues that are important for music enjoyment. Unfortunately, quantifying music enjoyment by CI users has been difficult because comparisons must rely on their recollection of music before they lost their hearing. Here, we aimed to assess music enjoyment in CI users using a readily interpretable reference based on acoustic hearing. The comparison was done by testing "single-sided deafness" (SSD) patients who have normal hearing (NH) in one ear and a CI in the other ear. The study also aimed to assess binaural musical enjoyment, with the reference being the experience of hearing with a single NH ear. Three experiments assessed the effect of adding different kinds of input to the second ear: electrical, vocoded, or unmodified. Design: In experiment 1, music enjoyment in SSD-CI users was investigated using a modified version of the MUSHRA (MUltiple Stimuli with Hidden Reference and Anchor) method. Listeners rated their enjoyment of song segments on a scale of 0 to 200, where 100 represented the enjoyment obtained from a song segment presented to the NH ear, 0 represented a highly degraded version of the same song segment presented to the same ear, and 200 represented enjoyment subjectively rated as twice as good as the 100 reference. Stimuli consisted of acoustic only, electric only, acoustic and electric, as well as other conditions with low pass filtered acoustic stimuli. Acoustic stimulation was provided by headphone to the NH ear and electric stimulation was provided by direct audio input to the subject's speech processor. In experiment 2, the task was repeated using NH listeners who received vocoded stimuli instead of electric stimuli. Experiment 3 tested the effect of adding the same unmodified song segment to the second ear, also in NH listeners. Results: Music presented through the CI only was very unpleasant, with an average rating of 20. Surprisingly, the combination of the unpleasant CI signal in one ear with acoustic stimulation in the other ear was rated more enjoyable (mean = 123) than acoustic processing alone. Presentation of the same monaural musical signal to both ears in NH listeners resulted with even greater enhancement of the experience compared with presentation to a single ear (mean = 159). Repeating the experiment using a vocoder to one ear of NH listeners resulted in interference rather than enhancement. Conclusions: Music enjoyment from electric stimulation is extremely poor relative to a readily interpretable NH baseline for CI-SSD listeners. However, the combination of this unenjoyable signal presented through a CI and an unmodified acoustic signal presented to a NH (or near-NH) contralateral ear results in enhanced music enjoyment with respect to the acoustic signal alone. Remarkably, this two-ear enhancement experienced by CI-SSD listeners represents a substantial fraction of the two-ear enhancement seen in NH listeners. This unexpected benefit of electroacoustic auditory stimulation will have to be considered in theoretical accounts of music enjoyment and may facilitate the quest to enhance music enjoyment in CI users.

© 2019 American Academy of Audiology. All rights reserved. Background: Frequency lowering (FL) technology offers a means of improving audibility of high-frequency sounds. For some listeners, the benefit of such technology can be accompanied by a perceived degradation in sound quality, depending on the strength of the FL setting. Purpose: The studies presented in this article investigate the effect of a new type of FL signal processing for hearing aids, adaptive nonlinear frequency compression (ANFC), on subjective speech quality. Research Design: Listener ratings of sound quality were collected for speech stimuli processed with systematically varied fitting parameters. Study Sample: Study 1 included 40 normal-hearing (NH) adult and child listeners. Study 2 included 11 hearing-impaired (HI) adult and child listeners. HI listeners were fitted with laboratory-worn hearing aids for use during listening tasks. Intervention: Speech quality ratings were assessed across test conditions consisting of various strengths of static nonlinear frequency compression (NFC) and ANFC speech. Test conditions included those that were fine-tuned on an individual basis per hearing aid fitting and conditions that were modified to intentionally alter the sound quality of the signal. Data Collection and Analysis: Listeners rated speech quality using the MUlti Stimulus test with Hidden Reference and Anchor (MUSHRA) test paradigm. Ratings were analyzed for reliability and to compare results across conditions. Results: Results show that interrater reliability is high for both studies, indicating that NH and HI listeners from both adult and child age groups can reliably complete the MUSHRA task. Results comparing sound quality ratings across experimental conditions suggest that both the NH and HI listener groups rate the stimuli intended to have poor sound quality (e.g., anchors and the strongest available parameter settings) as having below-average sound quality ratings. A different trend in the results is reported when considering the other experimental conditions across the listener groups in the studies. Speech quality ratings measured with NH listeners improve as the strength of ANFC decreases, with a range of bad to good ratings reported, on average. Speech quality ratings measured with HI listeners are similar and above-average for many of the experimental stimuli, including those with fine-tuned NFC and ANFC parameters. Conclusions: Overall, HI listeners provide similar sound quality ratings when comparing static and adaptive forms of frequency compression, especially when considering the individualized parameter settings. These findings suggest that a range in settings may result in above-average sound quality for adults and children with hearing impairment. Furthermore, the fitter should fine-tune FL parameters for each individual listener, regardless of type of FL technology.

Objectives Besides numerous other factors, listening experience with cochlear implants is substantially impaired by room acoustics. Even for persons without hearing impairment, the perception of auditory scenes, for example, concerning speech intelligibility, acoustic quality or audibility, is considerably influenced by room acoustics. For CI users, complex listening environments are usually associated with heavy losses. The aim of the present study was to determine room acoustic criteria that particularly influence speech pleasantness for CI users. Design Accordingly, speech material of the Oldenburg Sentence Test (Oldenburger Satztest, OLSA) as well as basic music material (major and minor triads) were auralized using the software Auratorium which allows auralization of simulated rooms. The constructed rooms for speech stimuli were based on the standard DIN 18041:2016–03 concerning acoustic quality in rooms, the binding standard referred to by room acoustic consultants in Germany, which also includes specifications for inclusive applications in schools. For the music perception tests, two typical concert halls of different sizes were modelled. The auralized test stimuli were unilaterally presented to 10 CI users via their auxiliary input as well as to 18 participants with typical hearing via headphones (control group). Speech pleasantness was evaluated using modified MUSHRA tests. Concerning music perception, chord discrimination was tested using paired comparisons. Results A strong preference of small source to listener distances by CI users was found, but no significant preference for room acoustic attenuation which exceeded the recommended for inclusive applications in schools. The analyses of the energy-time-structures suggested that a dense concentration of early reflections makes a beneficial impact on CI listeners’ pleasantness ratings. Music materials were distinguished more consistently without any room acoustic impact, while any room acoustic impact led to performance close to chance level. This effect is probably due to spectral smearing effects caused by reverberation. Conclusions These results suggest that in terms of pleasantness of speech, for CI-users, source-to-listener distance is the more influential parameter than room attenuation which goes beyond the German standard recommendation. Reflections from which CI users can benefit seem to occur much earlier than those from which NH listeners benefit. Future studies on chord discrimination concerning room acoustics are needed.

In this paper, a concealment method based on multiple-description coding (MDC) is presented, to improve speech quality deterioration caused by packet loss for algebraic code-excited linear prediction (ACELP) based coders. We apply to the ITU-T G.722.2 coder, a packet loss concealment (PLC) technique, which uses packetization schemes based on MDC. This latter is used with two new designed modes, which are modes 5 and 6 (18,25 and 19,85 kbps, respectively). We introduce our new second-order Markov chain model with four states in order to simulate network losses for different loss rates. The performance measures, with objective and subjective tests under various packet loss conditions, show a significant improvement of speech quality for ACELP based coders. The wideband perceptual evaluation of speech quality (WB-PESQ), enhanced modified bark spectral distortion (EMBSD), mean opinion score (MOS) tests and MUltiple Stimuli with Hidden Reference and Anchor (MUSHRA) for speech extracted from TIMIT database confirm the efficiency of our proposed approach and show a considerable enhancement in speech quality compared to the embedded algorithm in the standard ITU-T G.722.2.

Supplemental Digital Content is available in the text., Objectives: Cochlear implant (CI) users frequently report poor sound quality while listening to music, although the specific parameters responsible for this loss of sound quality remain poorly understood. Audio compression, which reduces the dynamic range (DR) for a given sound, is a ubiquitous component of signal processing used by both CI and hearing aid technology. However, the relative impact of compression for acoustic and electric hearing on music perception has not been well studied, an important consideration especially given that most compression algorithms in CIs were developed to optimize speech perception. The authors hypothesized that normal-hearing (NH) listeners would detect increased levels of compression more easily than CI users, but that both groups would perceive a loss of sound quality with increasing compression levels. Design: The present study utilizes the Cochlear Implant-MUltiple Stimulus with Hidden Reference and Anchor to evaluate the listener sensitivity to increasing levels of compression applied to music stimuli. The Cochlear Implant-MUltiple Stimulus with Hidden Reference and Anchor is a tool used to assess relative changes in the perceived sound quality of music across increasingly degraded listening conditions, in both CI and NH subjects. In this study, the authors applied multiple iterations of an aggressive compression algorithm to the music clips using Adobe Audition. The test conditions included 1, 3, 5, and 20 iterations sound tokens, with the 20-iteration samples serving as the Anchor stimuli. The compressed excerpts were 5 sec in length, with five clips for each of the five common musical genres (i.e., Classical, Jazz, Country, Rock, and Hip-Hop). Subjects were also presented with a Reference excerpt, which was the original music clip without any additional compression applied. CI recipients (n = 7, 11 ears) and NH listeners (n = 10) were asked to rate the sound quality of additionally compressed music as compared to the Reference. Results: Although both NH and CI groups could detect sound quality differences as a function of compression level, the discriminatory ability of the CI group was blunted compared to the NH group. The CI group had less variability in their responses and overall demonstrated reduced sensitivity to deterioration caused by excessive levels of compression. On average, the CI group rated the Anchor condition as only “Slightly worse” than the Reference. The music clips that were most affected by the compression were from Jazz and Hip-Hop genres and less so for Rock and Country clips. Corollary to this was a small but statistically significant impact of DR of the music clips on sound quality ratings, with narrower DR showing an association with poorer ratings. Conclusions: These results indicate that CI users exhibit less sensitivity to sound quality changes in music attributable to high levels of compression. These findings may account for another contributing factor to the generally poor music perception observed in CI users, particularly when listening to commercially recorded music.

Artificial speech synthesis has made a great leap in terms of naturalness as recent Text-to-Speech (TTS) systems are capable of producing speech with similar quality to human recordings. However, not all speaking styles are easy to model: highly expressive voices are still challenging even to recent TTS architectures since there seems to be a trade-off between expressiveness in a generated audio and its signal quality. In this paper, we present a set of techniques that can be leveraged to enhance the signal quality of a highly-expressive voice without the use of additional data. The proposed techniques include: tuning the autoregressive loop's granularity during training; using Generative Adversarial Networks in acoustic modelling; and the use of Variational Auto-Encoders in both the acoustic model and the neural vocoder. We show that, when combined, these techniques greatly closed the gap in perceived naturalness between the baseline system and recordings by 39% in terms of MUSHRA scores for an expressive celebrity voice., Comment: 6 pages, 4 figures, 2 tables, SSW 2021

In recent years, neural vocoders have surpassed classical speech generation approaches in naturalness and perceptual quality of the synthesized speech. Computationally heavy models like WaveNet and WaveGlow achieve best results, while lightweight GAN models, e.g. MelGAN and Parallel WaveGAN, remain inferior in terms of perceptual quality. We therefore propose StyleMelGAN, a lightweight neural vocoder allowing synthesis of high-fidelity speech with low computational complexity. StyleMelGAN employs temporal adaptive normalization to style a low-dimensional noise vector with the acoustic features of the target speech. For efficient training, multiple random-window discriminators adversarially evaluate the speech signal analyzed by a filter bank, with regularization provided by a multi-scale spectral reconstruction loss. The highly parallelizable speech generation is several times faster than real-time on CPUs and GPUs. MUSHRA and P.800 listening tests show that StyleMelGAN outperforms prior neural vocoders in copy-synthesis and Text-to-Speech scenarios., Accepted to ICASSP2021

In this article, we present a comprehensive scheme for the quality assessment of compressed vibrotactile signals with human assessors. Inspired by the multiple stimulus test with hidden reference and anchors (MUSHRA) from the audio domain, we designed a method in which each compressed signal is compared to its original signal and rated on a numerical scale. For each signal tested, the hidden reference and two anchor signals are used to validate the results and provide assessor screening criteria. Differing from previous approaches, our method is hierarchically structured and strictly timed in a sequential manner to avoid experimental confounds and provide precise psychophysical assessments. We validated our method in an experiment with 20 human participants in which we compared two state-of-the-art lossy codecs. The results show that, with our approach, the performance of different codecs can be compared effectively. Furthermore, the method also provides a measure of subjective quality at different data compression rates. The proposed procedure can be easily adapted to evaluate other vibrotactile codecs.

Ambisonics es un método para la reproducción inmersiva de audio espacial con ventajas técnicas relacionadas con la interactividad y la realidad virtual. Del mismo modo, también se conocen deficiencias en la resolución espacial cuando se utilizan grabaciones de primer orden. Utilizar órdenes superiores corrige estas deficiencias a costa de una mayor complejidad técnica y esfuerzo económico en los micrófonos que se utilizan. Cuando sólo se dispone de material en primer orden, pueden utilizarse estrategias de upmixing para aumentar la resolución espacial y el sweet spot o punto óptimo. Existen diferentes estrategias, y son estas las que se van a evaluar en este proyecto utilizando una esfera completa de 24 altavoces. Dado que las señales de los altavoces pueden convertirse en señales binaurales por medio de altavoces virtuales, las tres estrategias (DirAC, HARPEX y COMPASS) se comparan también reproduciendo las escenas a través de auriculares. El objetivo de este proyecto es presentar y analizar mediante una prueba de audio llevada a cabo en la UPNA los diferentes algoritmos de upmixing utilizados para convertir señales de Ambisonics de primer orden a señales Ambisonics de tercer orden por medio de rutinas de software. Graduado o Graduada en Ingeniería en Tecnologías de Telecomunicación por la Universidad Pública de Navarra Telekomunikazio Teknologien Ingeniaritzako Graduatua Nafarroako Unibertsitate Publikoan

In incremental text to speech synthesis (iTTS), the synthesizer produces an audio output before it has access to the entire input sentence. In this paper, we study the behavior of a neural sequence-to-sequence TTS system when used in an incremental mode, i.e. when generating speech output for token n, the system has access to n + k tokens from the text sequence. We first analyze the impact of this incremental policy on the evolution of the encoder representations of token n for different values of k (the lookahead parameter). The results show that, on average, tokens travel 88% of the way to their full context representation with a one-word lookahead and 94% after 2 words. We then investigate which text features are the most influential on the evolution towards the final representation using a random forest analysis. The results show that the most salient factors are related to token length. We finally evaluate the effects of lookahead k at the decoder level, using a MUSHRA listening test. This test shows results that contrast with the above high figures: speech synthesis quality obtained with 2 word-lookahead is significantly lower than the one obtained with the full sentence., Comment: 5 pages, 4 figures

OBJECTIVE: To determine sound quality for extended bandwidth amplification using a direct drive hearing device. STUDY DESIGN: Prospective double-blind within-subjects repeated measures study. SETTING: University hearing research laboratories. PATIENTS: Fifteen experienced hearing aid users with symmetric mild-sloping-to-severe sensorineural hearing loss. INTERVENTIONS: Sound quality ratings of speech and music passages were obtained using the Multiple Stimulus with Hidden References and Anchors (MUSHRA) protocol after wearing a direct drive hearing aid for at least 4 weeks. Passages were processed to filter out low-frequency (below 123 and 313 Hz) and high-frequency (above 4455, 5583, 6987, and 10,869 Hz) energy. MAIN OUTCOME MEASURES: Comparison of sound quality ratings for speech and music between low and high-pass filter frequencies measured from 0 to 100, where 0 represents "bad" and 100 represents "excellent." RESULTS: Wider bandwidth stimuli received higher sound quality ratings compared with narrower bandwidth stimuli. Conditions with more low-frequency energy (full-band and 123 Hz cut-off) were rated as having higher sound quality. More low-frequency energy in the 123 Hz condition was rated as having higher sound versus the 313 Hz condition (mean difference: 11.2%, p = 0.001). Full-band conditions with more low- and high-frequency energy were higher than the other high-frequency cutoff conditions (mean difference range: 12.9-15%, p

We propose a Text-to-Speech method to create an unseen expressive style using one utterance of expressive speech of around one second. Specifically, we enhance the disentanglement capabilities of a state-of-the-art sequence-to-sequence based system with a Variational AutoEncoder (VAE) and a Householder Flow. The proposed system provides a 22% KL-divergence reduction while jointly improving perceptual metrics over state-of-the-art. At synthesis time we use one example of expressive style as a reference input to the encoder for generating any text in the desired style. Perceptual MUSHRA evaluations show that we can create a voice with a 9% relative naturalness improvement over standard Neural Text-to-Speech, while also improving the perceived emotional intensity (59 compared to the 55 of neutral speech)., Accepted to ICASSP 2020

This paper presents a novel 3DoF+ system that allows to navigate, i.e., change position, in scene-based spatial audio content beyond the sweet spot of a Higher Order Ambisonics recording. It is one of the first such systems based on sound capturing at a single spatial position. The system uses a parametric decomposition of the recorded sound field. For the synthesis, only coarse distance information about the sources is needed as side information but not the exact number of them. In an experiment with multiple sources and reverberation, the proposed method shows good performance in terms of an instrumental measure and excellent performance in a subjective evaluation based on a MUSHRA listening test. A variant of the analysis stage increases localization performance and scores median subject ratings between 91 % and 99 % compared to the reference.

Sounds recorded by a dashboard camera presents essential evidence on a vehicle's driving condition for analyzing car accidents. In reality, when a direct impact of a vehicle equipped with a dashboard camera occurs, the power supply to the dashboard camera is abruptly terminated, and the digital video being recorded at that time may be incompletely stored; thus, there is no information such as the location or size to be read for the image and audio decoding. In this paper, we propose a method for recovering the linear pulse code modulation (PCM) audio from the impaired MP4 file partially stored in a video event data recorder (VEDR), known as a dashboard camera, during road vehicle accidents. This method is based on the principle of extracting an actual audio signal using the energy of each frame in the pseudo audio, which is generated from all data in the “mdat” area of the impaired MP4 file. For performance evaluation, the various video files with linear PCM audio format in the MP4 container were obtained through the different dashboard cameras. As the objective evaluation, we measured the audio recovery ratio, distance based on dynamic time warping (DTW) and comparison of the waveform and spectrogram. Moreover, the performance of the proposed method was evaluated using the multiple stimuli with a hidden reference and anchor (MUSHRA) test as the subjective assessment. As a result, it is shown from the evaluation that the proposed method can recover significantly better the linear PCM audio signal than the conventional methods from an impaired MP4 file.

Creating sound zones has been an active research field since the idea was first proposed. So far, most sound zone control methods rely on either an optimization of physical metrics such as acoustic contrast and signal distortion or a mode decomposition of the desired sound field. By using these types of methods, approximately 15 dB of acoustic contrast between the reproduced sound field in the target zone and its leakage to other zone(s) has been reported in practical set-ups, but this is typically not high enough to satisfy the people inside the zones. In this paper, we propose a sound zone control method shaping the leakage errors so that they are as inaudible as possible for a given acoustic contrast. The shaping of the leakage errors is performed by taking the time-varying input signal characteristics and the human auditory system into account when the loudspeaker control filters are calculated. We show how this shaping can be performed using variable span trade-off filters, and we show theoretically how these filters can be used for trading signal distortion in the target zone for acoustic contrast. The proposed method is evaluated based on physical metrics such as acoustic contrast and perceptual metrics such as STOI. The computational complexity and processing time of the proposed method for different system set-ups are also investigated. Lastly, the results of a MUSHRA listening test are reported. The test results show that the proposed method provides more than 20% perceptual improvement compared to existing sound zone control methods., Accepted for publication in IEEE/ACM TRANSACTIONS ON AUDIO, SPEECH, AND LANGUAGE PROCESSING

Phase serves as a critical component of speech that influences the quality and intelligibility. Current speech enhancement algorithms are beginning to address phase distortions, but the algorithms focus on normal-hearing (NH) listeners. It is not clear whether phase enhancement is beneficial for hearing-impaired (HI) listeners. We investigated the influence of phase distortion on speech quality through a listening study, in which NH and HI listeners provided speech-quality ratings using the MUSHRA procedure. In one set of conditions, the speech was mixed with babble noise at 4 different signal-to-noise ratios (SNRs) from -5 to 10 dB. In another set of conditions, the SNR was fixed at 10 dB and the noisy speech was presented in a simulated reverberant room with T60s ranging from 100 to 1000 ms. The speech level was kept at 65 dB SPL for NH listeners and amplification was applied for HI listeners to ensure audibility. Ideal ratio masking (IRM) was used to simulate speech enhancement. Two objective metrics (i.e., PESQ and HASQI) were utilized to compare subjective and objective ratings. Results indicate that phase distortion has a negative impact on perceived quality for both groups and PESQ is more closely correlated with human ratings., Comment: accepted by Interspeech2020, 5 pages, 4 figures

In this paper, the relationship between the positioning signal quality and sound quality at different mixture ratio was measured as a design criteria of acoustic indoor positioning system. In the experiment, the positioning signal quality was evaluated as peak-to-root-mean-square ratio (PR ratio) and the sound quality was evaluated by the MUSHRA score with 11 subjects. The experiments were performed for different signal-to-music ratio (SM ratio). As results, it was found that the SM ratio in a range of −40-0 (dB) affects the PR (ratio) and the listener feels uncomfortable with the SM ratio of higher than −26 (dB). The results of this experiment may provide criteria to design acoustic positioning signal in the audible range.

The paper presents the concept of an automatic system for music signal correction, considering room frequency response and music genre being played. The proposed algorithm, based on the room frequency response, compensates acoustic conditions surrounding the sound source. Additionally, the compensation process considers the signal content by recognizing music genre. As part of the described research, a series of subjective tests was presented to assess the effectiveness of the proposed algorithm depending on the music genre and acoustic conditions. The subjective tests described were carried out in accordance with the Mushra and AB methodology. In total, over 80 people took part in both tests and the results obtained from them were tested for statistical significance with the ANOVA test.

Large-scale mobile communication systems tend to contain legacy transmission channels with narrowband bottlenecks, resulting in characteristic "telephone-quality" audio. While higher quality codecs exist, due to the scale and heterogeneity of the networks, transmitting higher sample rate audio with modern high-quality audio codecs can be difficult in practice. This paper proposes an approach where a communication node can instead extend the bandwidth of a band-limited incoming speech signal that may have been passed through a low-rate codec. To this end, we propose a WaveNet-based model conditioned on a log-mel spectrogram representation of a bandwidth-constrained speech audio signal of 8 kHz and audio with artifacts from GSM full-rate (FR) compression to reconstruct the higher-resolution signal. In our experimental MUSHRA evaluation, we show that a model trained to upsample to 24kHz speech signals from audio passed through the 8kHz GSM-FR codec is able to reconstruct audio only slightly lower in quality to that of the Adaptive Multi-Rate Wideband audio codec (AMR-WB) codec at 16kHz, and closes around half the gap in perceptual quality between the original encoded signal and the original speech sampled at 24kHz. We further show that when the same model is passed 8kHz audio that has not been compressed, is able to again reconstruct audio of slightly better quality than 16kHz AMR-WB, in the same MUSHRA evaluation.

The following document tries out different statistical norms to validate the quality of synthesized voices applied to an HTS-based spanish synthesizer, which uses LSP and Cepstral Coefficients parameterizations. The standard MOS tests were carried out. Other types of quality tests were performed as well to reinforce the MOS results. Such as: MUSHRA, ABX and CCR. The subjective test PESQ was also applied. To validate intelligibility a SUS test was used.

In this paper, we investigate the coding efficiency of perceptual coding using an adaptive non-uniform orthogonal filter-bank based on MDCT analysis/synthesis and time domain aliasing reduction. We compare its performance to a system using a traditional adaptive uniform MDCT filterbank with window switching. The comparison is performed using a listening test at two different quantization settings. The statistical evaluation shows that the percetpual quality of the nonuniform filterbank significantly out-performs that of the uniform filterbank by 5 to 10 MUSHRA points.

This study proposes a new method of audio coding based on spectral recovery, which can enhance the performance of transform audio coding. An encoder represents spectral information of an input in a time-frequency domain and transmits only a portion of it so that the remaining spectral information can be recovered based on the transmitted information. A decoder recovers the magnitudes of missing spectral information using a convolutional neural network. The signs of missing spectral information are either transmitted or randomly assigned, according to their importance. By combining transmission and recovery of spectral information, the proposed method can enhance the coding performance, compared with conventional transform coding. The subjective performance evaluation shows that, for mono coding at 39.4 kbps, the proposed method provides higher sound quality than the USAC, by an average MUSHRA score of 8.5.

The perceived qualities of nine different single-microphone noise reduction (SMNR) algorithms were to be evaluated and compared in subjective listening tests with normal hearing and hearing impaired (HI) listeners.Speech samples added with traffic noise or with party noise were processed by the SMNR algorithms. Subjects rated the amount of speech distortions, intrusiveness of background noise, listening effort and overall quality, using a simplified MUSHRA (ITU-R, 2003 ) assessment method.18 normal hearing and 18 moderately HI subjects participated in the study.Significant differences between the rating behaviours of the two subject groups were observed: While normal hearing subjects clearly differentiated between different SMNR algorithms, HI subjects rated all processed signals very similarly. Moreover, HI subjects rated speech distortions of the unprocessed, noisier signals as being more severe than the distortions of the processed signals, in contrast to normal hearing subjects.It seems harder for HI listeners to distinguish between additive noise and speech distortions or/and they might have a different understanding of the term "speech distortion" than normal hearing listeners have. The findings confirm that the evaluation of SMNR schemes for hearing aids should always involve HI listeners.