Improving Speech Recognition Rate through Analysis Parameters

Eringis, Deividas; Tamulevičius, Gintautas

doi:10.2478/ecce-2014-0009

Cited by 14 publications

(5 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The first step for improving the recognition rate is to adjust the frame width and increment during feature extraction. This improvement has already been shown (Paliwal et al, 2010. Eringis & Tamulevicius, 2014 given that these parameters determine whether the modeling algorithms are getting a sufficient level of information from the speech segment as input.…”

Section: Tuning the Parameters For Feature Extractionmentioning

confidence: 79%

“…Similar conclusions were deducted (Gulzar et al 2014;Dave, 2013), but instead of adjusting the frame width, the effect of changing the set of features entirely was studied. In some other work (Eringis and Tamulevicius, 2014) it was shown that by adjusting the frame width and increment for different features, an improvement of 4.15% (from 88.75% to 92.9%) can be achieved.…”

Section: Introductionmentioning

confidence: 98%

See 1 more Smart Citation

Feature Fusion Based on Auditory and Speech Systems for an Improved Voice Biometrics System Using Artificial Neural Network

2020

14th International Conference on Computer Graphics, Visualization, Computer Vision and Image Processing, 5th International Conf

View full text Add to dashboard Cite

In today's world, identifying a speaker has become an essential task. Especially for systems that rely on voice commands or speech in general to operate. These systems use speaker-specific features to identify the individual, features such as Mel Frequency Cepstral Coefficients, Linear Predictive Coding, or Perceptual Linear Predictive. Although these features provide different representations of speech, they can all be considered as either auditory system based (type 1) or speech system based (type 2). In this work, a method for improving existing voice biometrics system is presented. A method fusing a type 1 feature with a type 2 feature is implemented using artificial neural network and evaluated on in-campus recorded data set. The obtained results demonstrate the potential of our approach in improving voice biometrics system, regardless of the underlying task being speaker identification or verification.

show abstract

Section: Tuning the Parameters For Feature Extractionmentioning

confidence: 79%

Section: Introductionmentioning

confidence: 98%

Feature Fusion Based on Auditory and Speech Systems for an Improved Voice Biometrics System Using Artificial Neural Network

2020

14th International Conference on Computer Graphics, Visualization, Computer Vision and Image Processing, 5th International Conf

View full text Add to dashboard Cite

show abstract

“…The formant extraction method is only reported in study A18. However, formant values may differ depending on the extraction method (e.g., linear predictive coding, Fast Fourier Transform, cepstral analysis), and on extraction/analysis parameters (such as window type, frame size, time step and parameters specific to each method) (Derdemezis et al, 2016;Eringis & Tamulevičius, 2014). Hence, this lack of information does not allow for the replication of the study's methodology, nor for comparative analyses.…”

Section: Consonant Measuresmentioning

confidence: 99%

Relationship between phoneme-level spectral acoustics and speech intelligibility in healthy speech: a systematic review

Pommée

Balaguer

Pinquier

et al. 2021

Speech, Language and Hearing

View full text Add to dashboard Cite

This study aims to systematically review original articles investigating the link between spectral acoustic measures in healthy talkers and perceived speech intelligibility, according to the PRISMA guidelines. Twenty-two studies were retained. Eighteen papers investigated vowel acoustics, one studied glides and eight articles investigated consonants, mostly sibilants. Various spectral measures and intelligibility estimates were used. The following measures were shown to be linked to sub-lexical perceived speech intelligibility ratings: for vowels, steady-state F1 and F2 measures, the F1 range, the [i]-[U] F2 difference, F0-F1 and F1-F2 differences in [è-A] and [q-è], the vowel space area, the mean amount of formant movement, the vector length and the spectral change measure; for consonants, the centroid energy and the spectral peak in the [s]-sound, as well as the steady-state F1 offset frequency in vowels preceding [t] and [d].To conclude, as speech is highly variable even in healthy adult speakers, a better understanding of the imprecisions in healthy spontaneous speech will provide a more realistic baseline for the investigation of disordered speech. To date, no acoustic measure is able to predict speech intelligibility to a large extent. There is still extensive research to be carried out to identify relevant acoustic combinations that could account for perceived speech variations (e.g. vowel and consonant reductions) and to gather normative data from a large number of healthy speakers. To that end, speech-related terms (e.g. intelligibility, comprehensibility, severity) need to be clearly defined and methodologies described in sufficient details to allow for replication, cross-comparisons/ meta-analyses and pooling of data. .

show abstract

“…Moreover, the phase information is estimated during the training that leads to better prediction of the clean signal, and also no scaling is needed for the output signal. However, working in the time domain results in a much higher number of network parameters due to the large frame size used, which is proved to be better than smaller frames [12], [26]. This larger number of parameters increases the size of the model, and restricts its applicability in some real time implementations as the model may not fit into the hardware [27].…”

Section: Literature Reviewmentioning

confidence: 99%

A Comparative Study of Time and Frequency Domain Approaches to Deep Learning based Speech Enhancement

Nossier

Wall

Moniri

et al. 2020

2020 International Joint Conference on Neural Networks (IJCNN)

View full text Add to dashboard Cite

Deep learning has recently made a breakthrough in the speech enhancement process. Some architectures are based on a time domain representation, while others operate in the frequency domain; however, the study and comparison of different networks working in time and frequency is not reported in the literature. In this paper, this comparison between time and frequency domain learning for five Deep Neural Network (DNN) based speech enhancement architectures is presented. The comparison covers the evaluation of the output speech using four objective evaluation metrics: PESQ, STOI, LSD, and SSNR increase. Furthermore, the complexity of the five networks was investigated by comparing the number of parameters and processing time for each architecture. Finally some of the factors that affect learning in time and frequency were discussed. The primary results of this paper show that fully connected based architectures generate speech with low overall perception when learning in the time domain. On the other hand, convolutional based designs give acceptable performance in both frequency and time domains. However, time domain implementations show an inferior generalization ability. Frequency domain based learning was proved to be better than time domain when the complex spectrogram is used in the training process. Additionally, feature extraction is also proved to be very effective in DNN based supervised speech enhancement, whether it is performed at the beginning, or implicitly by bottleneck layer features. Finally, it was concluded that the choice of the working domain is mainly restricted by the type and design of the architecture used.

show abstract

Improving Speech Recognition Rate through Analysis Parameters

Cited by 14 publications

References 33 publications

Feature Fusion Based on Auditory and Speech Systems for an Improved Voice Biometrics System Using Artificial Neural Network

Feature Fusion Based on Auditory and Speech Systems for an Improved Voice Biometrics System Using Artificial Neural Network

Relationship between phoneme-level spectral acoustics and speech intelligibility in healthy speech: a systematic review

A Comparative Study of Time and Frequency Domain Approaches to Deep Learning based Speech Enhancement

Contact Info

Product

Resources

About