Maximizing Phoneme Recognition Accuracy for Enhanced Speech Intelligibility in Noise

Petkov, Petko N.; Henter, Gustav Eje; Kleijn, W. Bastiaan

doi:10.1109/tasl.2013.2244089

Cited by 17 publications

(15 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…where we used Bayes' rule and whereâ L (u T , C), abbreviated toâ L , is the set of acoustic features observed by the listener, which is modeled as a deterministic function of the talker phoneme sequence u T and the speech modification parameters C. The first term of ( 16) is the likelihood of the talker phoneme sequence for the observed featuresâ L , the second term is the a-priori probability that the phoneme sequence u T is decoded by the listener, and the third term is the inverse a-priori probability of the listener observed features. Optimization of the likelihood term only reduces complexity and provides good results [11].…”

Section: B Intelligibility Enhancement Systemsmentioning

confidence: 99%

“…The optimization criteria vary widely as the signal processing algorithms are derived from different viewpoints and with different computational and delay constraints. Criteria used include the probability of correct phoneme recognition [11], auditory models [6], [13], [14], the articulation index [2], the speech intelligibility index [4], [8], mutual information [15], and sound-field distortion [16].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Optimizing speech intelligibility in a noisy environment: A unified view

Kleijn¹,

Crespo²,

Hendriks³

et al. 2021

Preprint

View full text Add to dashboard Cite

Modern communication technology facilitates communication from anywhere to anywhere. As a result, low speech intelligibility has become a common problem, which is exacerbated by the lack of feedback to the talker about the rendering environment. In recent years, a range of algorithms has been developed to enhance the intelligibility of speech rendered in a noisy environment. We describe methods for intelligibility enhancement from a unified vantage point. Before one defines a measure of intelligibility, the level of abstraction of the representation must be selected. For example, intelligibility can be measured on the message, the sequence of words spoken, the sequence of sounds, or a sequence of states of the auditory system. Natural measures of intelligibility defined at the message level are mutual information and the hit-or-miss criterion. The direct evaluation of high-level measures requires quantitative knowledge of human cognitive processing. Lower-level measures can be derived from higher-level measures by making restrictive assumptions. We discuss the implementation and performance of some specific enhancement systems in detail, including speech intelligibility index (SII)-based systems and systems aimed at enhancing the sound-field where it is perceived by the listener. We conclude with a discussion of the current state of the field and open problems. © 2015 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

show abstract

Section: B Intelligibility Enhancement Systemsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Optimizing speech intelligibility in a noisy environment: A unified view

Kleijn¹,

Crespo²,

Hendriks³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…phoneLLdscr: This entry builds on phoneLLabso, augmenting the objective measure with the difference of the measure in phoneLLabso, and the log of the sum of likelihoods conditioned on alternative acoustic models [26]. To reduce complexity, the context (phone neighbours) is assumed known and only a subset of all alternative models is considered based on the proximity of their LL scores to that obtained by the correct model.…”

Section: Challenge Entriesmentioning

confidence: 99%

Intelligibility-enhancing speech modifications: the hurricane challenge

Cooke¹,

Mayo²,

Valentini-Botinhao³

2013

Interspeech 2013

View full text Add to dashboard Cite

Speech output is used extensively, including in situations where correct message reception is threatened by adverse listening conditions. Recently, there has been a growing interest in algorithmic modifications that aim to increase the intelligibility of both natural and synthetic speech when presented in noise. The Hurricane Challenge is the first large-scale open evaluation of algorithms designed to enhance speech intelligibility. Eighteen systems operating on a common data set were subjected to extensive listening tests and compared to unmodified natural and text-to-speech (TTS) baselines. The best-performing systems achieved gains over unmodified natural speech of 4.4 and 5.1 dB in competing speaker and stationary noise respectively, while TTS systems made gains of 5.6 and 5.1 dB over their baseline. Surprisingly, for most conditions the largest gains were observed for noise-independent algorithms, suggesting that performance in this task can be further improved by exploiting information in the masking signal.

show abstract

“…The noise corrupted speech signal is of poor quality, as the noise components suppress the speech components and corrupts its cleanliness. On the other hand, speech intelligibility is the quantitative representation of speaker's intended message and the listener's perceived message [2]. It represents the understandability of the contents of the speech signal and it depends on the sound units, as the contents are defined by the sound units.…”

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

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

Maximizing Phoneme Recognition Accuracy for Enhanced Speech Intelligibility in Noise

Cited by 17 publications

References 23 publications

Optimizing speech intelligibility in a noisy environment: A unified view

Optimizing speech intelligibility in a noisy environment: A unified view

Intelligibility-enhancing speech modifications: the hurricane challenge

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

Contact Info

Product

Resources

About