GEDI: Gammachirp envelope distortion index for predicting intelligibility of enhanced speech

Yamamoto, Kazumasa; Araki, Shoko; Kinoshita, Keisuke; Nakatani, Tomohiro

doi:10.1016/j.specom.2020.06.001

Cited by 14 publications

(14 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The dcGC-FB was also used in models for speech intelligibility prediction [9][10][11][12][13]. A new model referred to as GEDI (the gammachirp envelope distortion index) [10,11] predicted the intelligibility of speech sounds processed with non-linear enhancement algorithms better than other recent indexes like STOI, CSII, and HASPI [12,13].…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

The gammachirp auditory filter and its application to speech perception

Patterson

2020

Acoust. Sci. & Tech.

Self Cite

View full text Add to dashboard Cite

We review the gammachirp (GC) auditory filter and its use in speech perception research. The GC was originally developed to explain the asymmetric, auditory filter shapes derived in notchednoise (NN) masking studies, and the strongly compressive input-output function observed in the mammalian cochlea. This compressive GC was fitted to a very large collection of notched-noise (NN) masking thresholds measured with a wide range of stimulus levels and center frequencies. The fit showed how the GC auditory filter could explain NN masking throughout the domain of human hearing with a relatively small number of parameters, only one of which was level dependent. Subsequently, a dynamic, compressive GC filterbank (dcGC-FB) was developed to simulate timedomain cochlear processing. This dcGC-FB has been used to cancel the peripheral compression of normal hearing and thereby simulate the most common forms of hearing loss. This simulator allows normal hearing listeners to experience the difficulties of hearing impaired listeners. It has been used in training courses for speech-language-hearing therapists and psychoacoustic experiments. The dcGC-FB has also been used for modeling speaker size perception and predicting speech intelligibility with GEDI (the gammachirp envelope distortion index).

show abstract

Section: Resultsmentioning

confidence: 99%

“…The GC architecture enables us to construct a hearing impairment simulator which allows normal hearing listeners to experience the difficulties of hearing impaired listeners [5,6]. The GC has also been used to model speaker size perception [7,8] and speech intelligibility [9][10][11][12][13].…”

Section: Introductionmentioning

confidence: 99%

The gammachirp auditory filter and its application to speech perception

Patterson

2020

Acoust. Sci. & Tech.

Self Cite

View full text Add to dashboard Cite

show abstract

“…The short time objective intelligibility (STOI) [3] computes average correlation coefficients over all the short-time frames of the 1/3-octave frequency bands. The speech intelligibility was derived as a percentage using a logistic function with the optimized parameters, which are described in [30]. The STOI measure correlated well with intelligibility of speech signals processed by the ideal binary mask algorithm [31].…”

Section: Stoimentioning

confidence: 99%

“…The speech intelligibility percentage can also be derived using a logistic function of the coherence and the cross-correlation. The optimized parameter values are described in [30].…”

Section: Haspimentioning

confidence: 99%

See 1 more Smart Citation

Predicting Speech Intelligibility of Enhanced Speech Using Phone Accuracy of DNN-Based ASR System

Arai¹,

Araki²,

Ogawa³

et al. 2019

Interspeech 2019

Self Cite

View full text Add to dashboard Cite

The ability of state-of-the-art automatic speech recognition (ASR) systems, which use deep neural networks (DNN), has recently been approaching that of human auditory systems. On the other hand, although measuring the intelligibility of enhanced speech signals is important for developing auditory algorithms and devices, the current measurement methods mainly rely on subjective experiments. Therefore, it would be preferable to employ an ASR system to predict the subjective speech intelligibility (SI) of enhanced speech. In this study, we evaluate the SI prediction performance of DNN-based ASR systems using phone accuracies. We found that an ASR system with multicondition training achieves the best SI prediction accuracy for enhanced speech when compared with conventional methods (STOI, HASPI) and a recently proposed technique (GEDI). In addition, since our ASR system uses only a phone language model, it offers the advantage of being able to predict intelligibility independently of prior knowledge of words.

show abstract