Enhanced Non-Intrusive Speech Quality Measurement Using Degradation Models

Falk, Tiago H.; Chan, Wai-Yip

doi:10.1109/icassp.2006.1660151

Cited by 8 publications

(7 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Current speech quality measurement algorithms should handle such degradation conditions efficiently. In previous work [24], estimating multiplicative noise is shown to be beneficial for GMM-based speech quality measurement. A multiplicative noise estimator, similar to the one described in [16], was deployed and performance improvement was reported for MNRU degradations.…”

Section: Detecting and Estimating Multiplicative Noisementioning

confidence: 99%

“…Here, a modification to the GMM-based architecture described in [13] is proposed. It is found that accuracy can be enhanced if the algorithm is also equipped with information regarding the behavior of speech degraded by different transmission and coding schemes [24]. To this end, clean speech signals are used to train three different Gaussian mixture densities, .…”

Section: Consistency Calculation and Mos Mappingmentioning

confidence: 99%

See 1 more Smart Citation

Single-Ended Speech Quality Measurement Using Machine Learning Methods

Falk

Chan

2006

IEEE Trans. Audio Speech Lang. Process.

Self Cite

View full text Add to dashboard Cite

Abstract-We describe a novel single-ended algorithm constructed from models of speech signals, including clean and degraded speech, and speech corrupted by multiplicative noise and temporal discontinuities. Machine learning methods are used to design the models, including Gaussian mixture models, support vector machines, and random forest classifiers. Estimates of the subjective mean opinion score (MOS) generated by the models are combined using hard or soft decisions generated by a classifier which has learned to match the input signal with the models. Test results show the algorithm outperforming ITU-T P.563, the current "state-of-art" standard single-ended algorithm. Employed in a distributed double-ended measurement configuration, the proposed algorithm is found to be more effective than P.563 in assessing the quality of noise reduction systems and can provide a functionality not available with P.862 PESQ, the current double-ended standard algorithm.Index Terms-Mean opinion score (MOS), objective quality measurement, quality model, single-ended measurement, speech communication, speech distortions, speech enhancement, speech quality, subjective quality.

show abstract

Section: Detecting and Estimating Multiplicative Noisementioning

confidence: 99%

Section: Consistency Calculation and Mos Mappingmentioning

confidence: 99%

Single-Ended Speech Quality Measurement Using Machine Learning Methods

Falk

Chan

2006

IEEE Trans. Audio Speech Lang. Process.

Self Cite

View full text Add to dashboard Cite

show abstract

“…The noise PSD matrix (estimated during silent periods) and the diffuse sound power were estimated by an autoregressive averaging procedure with a time constant of 50 ms. The VM signal was evaluated using the average fullband segmental signal-to-diffuse-plus-noise ratio (SDNR), the speech-toreverberation modulation ratio (SRMR) [11], and PESQ score [12]. Fig.…”

Section: Setup and Performance Measuresmentioning

confidence: 99%

Sound acquisition in noisy and reverberant environments using virtual microphones

Kowalczyk

Thiergart

Crăciun

et al. 2013

2013 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics

View full text Add to dashboard Cite

In hands-free communication applications, the main goal is to capture desired sounds, while reducing noise and interfering sounds. However, for natural-sounding telepresence systems, the spatial sound image should also be preserved. Using a recently proposed method for generating the signal of a virtual microphone (VM), one can recreate the sound image from an arbitrary point of view in the sound scene (e.g., close to a desired speaker), while being able to place the physical microphones outside the sound scene. In this paper, we present a method for synthesizing a VM signal in noisy and reverberant environments, where the estimation of the required direct and diffuse sound components is performed using two multichannel linear filters. The direct sound component is estimated using a multichannel Wiener filter, while the diffuse sound component is estimated using a linearly constrained minimum variance filter followed by a single-channel Wiener filter. Simulations in a noisy and reverberant environment show the applicability of the proposed method for sound acquisition in a scenario in which two microphone arrays are installed in a large TV

show abstract

“…Although there are many studies which shows that the performance provided by this metric in VoIP service can be improved [5], [6] and [7]. This is due to the intrinsic characteristics of the algorithm that implements the P.563 metric, which is mainly based on the analysis of the vocal tract [4], and is not planned to consider external factors, such as, packet losses in an IP network.…”

Section: Introductionmentioning

confidence: 99%

Improving the performance of a non-intrusive metric of voice quality assessment considering IP network parameters

Pereira¹,

Nunes²,

Rosa³

et al. 2015

Anais De XXXIII Simpósio Brasileiro De Telecomunicações

View full text Add to dashboard Cite

In a voice-over-IP (VoIP) communication there are different factors that degrade the voice signal quality, and these degradations need to be evaluated in real time. In this scenario, the non-intrusive metrics such as the ITU-T P.563 recommendation are the most indicated. However, this metric has not a suitable performance if compared with intrusive metrics. In this study, an adjustment function is proposed to improve the P.563 metric's performance considering the packet loss parameter. The results show that the performance in relation to ITU-T P.862 was improved, obtaining a Pearson-correlationcoefficient and a maximum error of 0.99 and 0.40, respectively.

show abstract

Enhanced Non-Intrusive Speech Quality Measurement Using Degradation Models

Cited by 8 publications

References 7 publications

Single-Ended Speech Quality Measurement Using Machine Learning Methods

Single-Ended Speech Quality Measurement Using Machine Learning Methods

Sound acquisition in noisy and reverberant environments using virtual microphones

Improving the performance of a non-intrusive metric of voice quality assessment considering IP network parameters

Contact Info

Product

Resources

About