Near and Far Field Speech-in-Noise Intelligibility Improvements Based on a Time–Frequency Energy Reallocation Approach

Zorilă, Tudor-Cătălin; Stylianou, Yannis; Ishihara, Tatsuma

doi:10.1109/taslp.2016.2585864

Cited by 13 publications

(11 citation statements)

References 55 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…H1 sharpened the formants, H2 flattened the spectral tilt, and H3 preemphasized the mid frequency range. A similar voicing index as in [11] was used to adapt the first two operations. The last two stages of MBSS have performed SNR-adaptive gain thresholding and fixed time smoothing.…”

Section: Multiband Spectral Shapingmentioning

confidence: 99%

“…The first stage had four steps. One was reconstructing the waveform from the spectral coefficients; another one was applying full wave rectification on the previous waveform, extracting 30-ms frames at every 30-ms, picking the maximum value in each frame, and re-sampling the sequence of maxima at the original frame rate of MBSSDRC; at a third stage, the level of the envelope e(m) computed previously was statically compressed using the strategy described in [11], yielding the gains g(m); finally, the dynamics of g(m) were enhanced to have a faster response to on-sets and a slower one to off-sets,…”

Section: Multiband Dynamic Range Compressionmentioning

confidence: 99%

“…The standard SSDRC was designed to process clean speech [11] however, in realistic conditions, the input is often noisy. Although typical SNR levels (e.g., 10-20 dB) do not degrade intelligibility, they can have a significant effect on the perceptual quality of speech processed for NLE, particularly in the absence of near-end noise.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

On the Quality and Intelligibility of Noisy Speech Processed for Near-End Listening Enhancement

Zorilă¹,

Stylianou²

2017

Interspeech 2017

Self Cite

View full text Add to dashboard Cite

Most current techniques for near-end speech intelligibility enhancement have focused on processing clean input signals, however, in realistic environments, the input is often noisy. Processing noisy speech for intelligibility enhancement using algorithms developed for clean signals can lower the perceptual quality of the samples when they are listened in quiet. Here we address the quality loss in these conditions by combining noise reduction with a multi-band version of a state-of-the-art intelligibility enhancer for clean speech that is based on spectral shaping and dynamic range compression (SSDRC). Subjective quality and intelligibility assessments with noisy input speech showed that: (a) In quiet near-end conditions, the proposed system outperformed the baseline SSDRC in terms of Mean Opinion Score (MOS); (b) In speech-shaped near-end noise, the proposed system improved the intelligibility of unprocessed speech by a factor larger than three at the lowest tested signalto-noise ratio (SNR) however, overall, it yielded lower recognition scores than the standard SSDRC.

show abstract

Section: Multiband Spectral Shapingmentioning

confidence: 99%

Section: Multiband Dynamic Range Compressionmentioning

confidence: 99%

See 1 more Smart Citation

On the Quality and Intelligibility of Noisy Speech Processed for Near-End Listening Enhancement

Zorilă¹,

Stylianou²

2017

Interspeech 2017

Self Cite

View full text Add to dashboard Cite

show abstract

“…The first algorithm (SSDRC) was based on the work of Zorilȃ et al [4,7] and used a two-stage energy reallocation strategy. During the first stage (spectral shaping -SS), the speechto-noise ratio (SNR) at medium and high frequencies was increased by transferring energy from below 500 Hz to higher frequencies.…”

Section: Intelligibility Enhancement Algorithms For Evaluationmentioning

confidence: 99%

“…This problem is known as near-end listening enhancement or speech intelligibility enhancement (SINE) and the constraint is called the equal-RMS (EQR) constraint. Typical solutions use modifications that are either based on previous intelligibility studies (e.g., simulating the differences between Lombard speech and conversational speech or emphasis of the information-bearing segments of speech) [3][4][5][6][7] or they optimize an objective measure that correlates well with measured intelligibility [8][9][10][11]. 'Inclusive' algorithms that would work for both normal-hearing and hearing-impaired listeners or in reverberant conditions are becoming increasingly popular [12,13].…”

Section: Introductionmentioning

confidence: 99%

Effectiveness of Near-End Speech Enhancement Under Equal-Loudness and Equal-Level Constraints

et al. 2016

Self Cite

View full text Add to dashboard Cite

Most recently proposed near-end speech enhancement methods have been evaluated with the overall power (RMS) of the speech held constant. While significant intelligibility gains have been reported in various noisy conditions, an equal-RMS constraint may lead to enhancement solutions that increase the loudness of the original speech. Comparable effects might be produced simply by increasing the power of the original speech, which also leads to an increase in loudness. Here we suggest modifying the equal-RMS constraint to one of equal loudness between the original and the modified signals, based on a loudness model for time-varying sounds. Four state-of-the-art speechin-noise intelligibility enhancement systems were evaluated under the equal-loudness constraint, using intelligibility tests with normal-hearing listeners. Results were compared with those obtained under the equal-RMS constraint. The methods based on spectral shaping and dynamic range compression yielded significant intelligibility gains regardless of the constraint, while for the method without dynamic range compression the intelligibility gain was lower under the equal-loudness than under the equal-RMS constraint.

show abstract

A novel double pole transfer function‐single frequency filtering approach for speech enhancement

Srinivasarao

Ghanekar

2020

Trans Emerging Tel Tech

View full text Add to dashboard Cite

Speech intelligibility improvement is a major field of recent research with in audiology and hearing domain. Speech signal usually gets affected due to various environmental and background noise. Several approaches for speech intelligibility improvement were presented in the literature. In this paper, a method of speech enhancement based on Double Pole Transfer Function-Single Frequency Filter (DPTF-SFF) has been proposed. In the SFF approach the temporal and spectral resolutions are controlled with a one particular value of the filter that corresponds to the pole position in the complex z-plane. The proposed method uses double pole transfer function SFF that improves the magnitude and phase for various frequencies which are used in the synthesis of the original signal. In the analysis of this method clean speech signals are used from GRID corpus database and noise signals are taken from NOIZEUS and NOISEX-92 databases. The performance evaluation of this method is carried out in terms of Perceptual Evaluation of Speech Quality, Short-Time Objective Intelligibility, and Segmental Signal to Noise Ratio metrics. The implementation results proved that the proposed method is an efficient one as it gives higher intelligibility scores compared to the existing methods at different SNR levels (dB) and for different noises employed.

show abstract

Near and Far Field Speech-in-Noise Intelligibility Improvements Based on a Time–Frequency Energy Reallocation Approach

Cited by 13 publications

References 55 publications

On the Quality and Intelligibility of Noisy Speech Processed for Near-End Listening Enhancement

On the Quality and Intelligibility of Noisy Speech Processed for Near-End Listening Enhancement

Effectiveness of Near-End Speech Enhancement Under Equal-Loudness and Equal-Level Constraints

A novel double pole transfer function‐single frequency filtering approach for speech enhancement

Contact Info

Product

Resources

About