Audio signal quality enhancement using multi-layered convolutional neural network based auto encoder–decoder

Raj, Shivangi; Periasamy, P.; Gupta, Shubham

doi:10.1007/s10772-021-09809-z

Cited by 4 publications

(3 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

Section: Feature Extraction Using W-wpccmentioning

confidence: 98%

“…The following are the primary reasons why Inception-v3 is so effective: Inception-v3 has fewer than half (60,000,000) of the parameters of AlexNet and less than a quarter of the parameters of AlexNet. (140,000,000); also, the total amount of floating-point calculations performed by the complete Inception-v3 network is around 5,000,000,000 times, significantly greater than that of Inception-v1 (approximately 1,500,000,000 times)[28][29][30].These features enable Inception-v3 to be easier to implement, as it can be readily deployed on standard servers to deliver quick response services. Inception-v3 employs convolution kernels of varying sizes, resulting in various receptive fields.…”

mentioning

confidence: 99%

See 1 more Smart Citation

An Efficient Voice Authentication System using Enhanced Inceptionv3 Algorithm

2023

JMC

View full text Add to dashboard Cite

Automatic voice authentication based on deep learning is a promising technology that has received much attention from academia and industry. It has proven to be effective in a variety of applications, including biometric access control systems. Using biometric data in such systems is difficult, particularly in a centralized setting. It introduces numerous risks, such as information disclosure, unreliability, security, privacy, etc. Voice authentication systems are becoming increasingly important in solving these issues. This is especially true if the device relies on voice commands from the user. This work investigates the development of a text-independent voice authentication system. The spatial features of the voiceprint (corresponding to the speech spectrum) are present in the speech signal as a result of the spectrogram, and the weighted wavelet packet cepstral coefficients (W-WPCC) are effective for spatial feature extraction (corresponding to the speech spectrum). W- WPCC characteristics are calculated by combining sub-band energies with sub-band spectral centroids using a weighting scheme to generate noise-resistant acoustic characteristics. In addition, this work proposes an enhanced inception v3 model for voice authentication. The proposed InceptionV3 system extracts feature from input data from the convolutional and pooling layers. By employing fewer parameters, this architecture reduces the complexity of the convolution process while increasing learning speed. Following model training, the enhanced Inception v3 model classifies audio samples as authenticated or not based on extracted features. Experiments were carried out on the speech of five English speakers whose voices were collected from YouTube. The results reveal that the suggested improved method, based on enhanced Inception v3 and trained on speech spectrogram pictures, outperforms the existing methods. The approach generates tests with an average categorization accuracy of 99%. Compared to the performance of these network models on the given dataset, the proposed enhanced Inception v3 network model achieves the best results regarding model training time, recognition accuracy, and stability.

show abstract

Section: Feature Extraction Using W-wpccmentioning

confidence: 98%

mentioning

confidence: 99%

An Efficient Voice Authentication System using Enhanced Inceptionv3 Algorithm

2023

JMC

View full text Add to dashboard Cite

show abstract

“…As the end of the encoder, the input of the sequence with the beginning and end characters of the mother tongue is converted into a vector file, which transmits to the neural network, including the vector return of the input language file neural network. e end of the decoder, the neural network, is used as a carrier to calculate and access the target language temporarily [20].…”

Section: Encoder Decodermentioning

confidence: 99%

Design of English Translation Mobile Information System Based on Recurrent Neural Network

Gao¹

2022

Mobile Information Systems

View full text Add to dashboard Cite

To solve the problem of translating lines of difference in length into English, this article presents a model of neural network recovery (RNN) English translator-based models of end-to-end encoder-decoder. This method promotes machine autonomous learning of features and transforms corpus data into word vectors by constructing end-to-end. By mapping the source language and target language directly through the recurrent neural network and selecting semantic error to construct objective function during training, the influence of each part in semantic can be well balanced, and the alignment information is fully considered, which provides powerful guidance for deep recurrent neural network training. The results of the neural network test define the standard BLEU score by 1.51–11.86. Our test scores and BLEU scores at all levels show that data in equivalence play an important role in modeling. Summary. the English translation model based on the neural repetitive fusion is efficient and stable.

show abstract