An analysis of the influence of deep neural network (DNN) topology in bottleneck feature based language recognition

Lozano-Díez, Alicia; Zazo, Rubén; Toledano, Doroteo T.; González-Rodríguez, Joaquín

doi:10.1371/journal.pone.0182580

Cited by 51 publications

(29 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Finally the output layer O produces the output of the DNN for the target task (for the case of classification, the posterior probability of an input vector to belong to each of the C classes). Reprinted from [ 19 ] under a CC BY license, with permission from Alicia Lozano et. al., original copyright 2017.…”

Section: Deep Neural Network and Speech Processingmentioning

confidence: 99%

Multi-resolution speech analysis for automatic speech recognition using deep neural networks: Experiments on TIMIT

2018

Self Cite

View full text Add to dashboard Cite

Speech Analysis for Automatic Speech Recognition (ASR) systems typically starts with a Short-Time Fourier Transform (STFT) that implies selecting a fixed point in the time-frequency resolution trade-off. This approach, combined with a Mel-frequency scaled filterbank and a Discrete Cosine Transform give rise to the Mel-Frequency Cepstral Coefficients (MFCC), which have been the most common speech features in speech processing for the last decades. These features were particularly well suited for the previous Hidden Markov Models/Gaussian Mixture Models (HMM/GMM) state of the art in ASR. In particular they produced highly uncorrelated features of small dimensionality (typically 13 coefficients plus deltas and double deltas), which was very convenient for diagonal covariance GMMs, for dealing with the curse of dimensionality and for the limited computing resources of a decade ago. Currently most ASR systems use Deep Neural Networks (DNN) instead of the GMMs for modeling the acoustic features, which provides more flexibility regarding the definition of the features. In particular, acoustic features can be highly correlated and can be much larger in size because the DNNs are very powerful at processing high-dimensionality inputs. Also, the computing hardware has reached a level of evolution that makes computational cost in speech processing a less relevant issue. In this context we have decided to revisit the problem of the time-frequency resolution in speech analysis, and in particular to check if multi-resolution speech analysis (both in time and frequency) can be helpful in improving acoustic modeling using DNNs. Our experiments start with several Kaldi baseline system for the well known TIMIT corpus and modify them by adding multi-resolution speech representations by concatenating different spectra computed using different time-frequency resolutions and different post-processed and speaker-adapted features using different time-frequency resolutions. Our experiments show that using a multi-resolution speech representation tends to improve over results using the baseline single resolution speech representation, which seems to confirm our main hypothesis. However, results combining multi-resolution with the highly post-processed and speaker-adapted features, which provide the best results in Kaldi for TIMIT, yield only very modest improvements.

show abstract

Section: Deep Neural Network and Speech Processingmentioning

confidence: 99%

Multi-resolution speech analysis for automatic speech recognition using deep neural networks: Experiments on TIMIT

2018

Self Cite

View full text Add to dashboard Cite

show abstract

“…to the weighted sum and sends the result out of the neuron . The parameters W j (weights matrix) and b j (bias vectors) control the behavior of the DNN models by being adjusted repeatedly until the cost function reaches its minimum value [11].…”

Section: Deep Neural Network (Dnn)mentioning

confidence: 99%

A DNN-based Image Retrieval Approach for Detection of Defective Area in Carbon Fiber Reinforced Polymers through LDV Data

Basiri

Hasanzadeh

Hadi

et al. 2020

2020 International Conference on Machine Vision and Image Processing (MVIP)

View full text Add to dashboard Cite

Carbon fiber reinforced polymer (CFRP) materials, due to their specific strength and high consistency against erosion and corrosion, are widely used in industrial applications and high-tech engineering structures. However, there are also disadvantages: e.g. they are prone to different kinds of internal defects which could jeopardize the structural integrity of the CFRP material and therefore early detection of such defects can be an important task. Recently, local defect resonance (LDR), which is a subcategory of ultrasonic nondestructive testing, has been successfully used to solve this issue. However, the drawback of utilizing this technique is that the frequency at which the LDR occurs must be known. Further, the LDR-based technique has difficulty in assessing deep defects. In this paper, deep neural network (DNN) methodology is employed to remove this limitation and to acquire a better defect image retrieval process and also to achieve a model for the approximate depth estimation of such defects. In this regards, two types of defects called flat bottom holes (FBH) and barely visible impact damage (BVID) which are made in two CFRP coupons are used to evaluate the ability of the proposed method. Then, these two CFRPs are excited with a piezoelectric patch, and their corresponding laser Doppler vibrometry (LDV) response is collected through a scanning laser Doppler vibrometer (SLDV). Eventually, the superiority of our DNN-based approach is evaluated in comparison with other well-known classification methodologies.

show abstract

“…Despite the success of BNFs for SID [13,14,15] and LID [16,17,18,19,20,21], the variable length of this frame-wise representation poses a challenge in consequent modeling. The classical i-vector compacts the utterance representation in a fixedlength vector.…”

Section: Utterance Level Representation: Dnn-based Embeddingsmentioning

confidence: 99%

Bottleneck and Embedding Representation of Speech for DNN-based Language and Speaker Recognition

Lozano-Díez¹,

González-Rodríguez²,

Gónzalez-Domínguez³

2018

IberSPEECH 2018

Self Cite

View full text Add to dashboard Cite

In this manuscript, we summarize the findings presented in Alicia Lozano Diez's Ph.D. Thesis, defended on the 22nd of June, 2018 in Universidad Autonoma de Madrid (Spain). In particular, this Ph.D. Thesis explores different approaches to the tasks of language and speaker recognition, focusing on systems where deep neural networks (DNNs) become part of traditional pipelines, replacing some stages or the whole system itself. First, we present a DNN as classifier for the task of language recognition. Second, we analyze the use of DNNs for feature extraction at frame-level, the so-called bottleneck features, for both language and speaker recognition. Finally, utterancelevel representation of the speech segments learned by the DNN (known as embedding) is described and presented for the task of language recognition. All these approaches provide alternatives to classical language and speaker recognition systems based on i-vectors (Total Variability modeling) over acoustic features (MFCCs, for instance). Moreover, they usually yield better results in terms of performance.

show abstract

An analysis of the influence of deep neural network (DNN) topology in bottleneck feature based language recognition

Cited by 51 publications

References 13 publications

Multi-resolution speech analysis for automatic speech recognition using deep neural networks: Experiments on TIMIT

Multi-resolution speech analysis for automatic speech recognition using deep neural networks: Experiments on TIMIT

A DNN-based Image Retrieval Approach for Detection of Defective Area in Carbon Fiber Reinforced Polymers through LDV Data

Bottleneck and Embedding Representation of Speech for DNN-based Language and Speaker Recognition

Contact Info

Product

Resources

About