A Frequency-Domain Adaptive Line Enhancer With Step-Size Control Based on Mutual Information for Harmonic Noise Reduction

Taghia, Jalal; Martin, Rainer

doi:10.1109/taslp.2016.2545920

Cited by 18 publications

(9 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Common examples for harmonic noise in real‐life scenario include traffic noise, electronic noise, and so on, Harmonic noise samples are taken from the free sound effects [31] and sound ideas database [32]. The FDALE algorithm [24] with an adaptive step‐size control, proposed for harmonic noise reduction outperforms the FDALE algorithm (which uses a fixed step‐size control) [33], in terms of instrumental speech quality and intelligibility. The performance of the FDALE algorithm with adaptive step size is compared with our proposed techniques using PESQ and STOI measures and the results are given in Table 6.…”

Section: Performance Evaluation Of Sisquimentioning

confidence: 99%

“…Therefore, for further improvement in quality, improved noise removal in the noise dominant sub-bands than the speech dominant sub-bands is necessary that makes the noise removal to be adaptive. Adaptive noise removal techniques make use of minimum mean square error based spectral speech enhancement on the time-frequency (TF) unit and apply adaptive masking on the noise removed TF unit or exploits mutual information to derive the frequency dependent step size [23,24]. In the current work, modified spectral subtractionbased speech enhancement, proposed by the authors, is adopted on a dynamically (adaptively) varying multi-band signal for effective speech enhancement.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Adaptive multi‐band filter structure‐based far‐end speech enhancement

Jeeva

Nagarajan²,

Vijayalakshmi³

2020

IET signal process.

View full text Add to dashboard Cite

Speech signals degraded by noise tend to lose their quality and intelligibility. Therefore, the goal of speech enhancement algorithms is to restore these attributes of speech. The current work proposes a dynamic filter structure, dynamic over every utterance, that would vary simultaneously based on (i) the class of sound units in a given noisy/degraded signal, to improve intelligibility, and (ii) the noise components present, to improve quality. This filter structure is employed in the temporaldomain filtering-based multi-band speech enhancement algorithm, proposed by Jeeva et al.. The performance of the algorithm is evaluated subjectively and objectively, in terms of quality and intelligibility, and the algorithm is observed to successfully improve both attributes of degraded speech. Since the improvement in intelligibility depends on the effective restoration of sound units in the utterance, this process is language-specific. In this regard, an analysis is performed to study the influence of phone class distribution on the intelligibility improvement achieved for four Indian languages, namely Tamil, Hindi, Telugu, and Malayalam, and Indian English.

show abstract

Section: Performance Evaluation Of Sisquimentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Adaptive multi‐band filter structure‐based far‐end speech enhancement

Jeeva

Nagarajan²,

Vijayalakshmi³

2020

IET signal process.

View full text Add to dashboard Cite

show abstract

“…ALEs make use of the correlation differences between the tonals and the wideband noise components [2] and are considered to be important applications of adaptive filtering techniques [13,14]. Aside from passive sonar application, ALEs have also been used in speech enhancement [15,16] and biomedical signal processing [13,17]. As pointed out in the references [18,19], ALEs require that their input signal-to-noise ratios (SNRs) should be higher than certain thresholds.…”

Section: Introductionmentioning

confidence: 99%

Deep‐learning‐based line enhancer for passive sonar systems

Chi

et al. 2021

IET Radar Sonar & Navi

View full text Add to dashboard Cite

Detection of acoustic tonals from surfaces and underwater vehicles is important for passive sonar systems. Enhancements of the tonals are usually necessary in passive sonar prior to detections. Conventionally, passive sonars employ adaptive line enhancers (ALE) in order to realise enhancements of the tonals. However, ALEs have requirements on their input signalto-noise ratios (SNR). When the SNR inputs are too low, the ALEs cannot perform well. Therefore, for the purpose of overcoming the limitations of the SNR inputs to ALEs, this study proposes to enhance tonals using unsupervised deep-learning techniques. The proposed deep-learning-based line enhancer (DLE) is based on an autoencoder neural network. The simulation results show that when the input SNR is −28 dB, the proposed DLE still achieves an SNR gain of 15 dB. However, the reference ALE fails. The experimental results also demonstrate the superiority of the proposed DLE.

show abstract

“…To enhance the narrowband discrete components, passive sonars usually employ an adaptive line enhancer (ALE) as a pre‐processing step [9–11]. As an important application of the adaptive filter technique, ALEs have been used in many fields such as speech enhancement [12, 13] and biomedical signal processing [14]. The basic idea behind ALEs is to utilise the difference between the correlation lengths of the narrowband discrete components and wideband noise.…”

Section: Introductionmentioning

confidence: 99%