LSTM Technique Research Articles

Improving speech recognition using bionic wavelet features

Bionic wavelet transform is a continuous wavelet, based on adaptive time frequency technique. This paper presents a speech recognition system for recognizing isolated words by discretizing the continuous Bionic Wavelet (BW). Conversion from continuous to discrete is achieved by adopting central frequency and thresholding techniques. The BW features of noisy signal are processed through MFCC to obtain the optimal features of the speech signal. SVM, Artificial Neural Network (ANN) and LSTM techniques are used to improve the recognition rate by enhancing the speech signals. The experiments are conducted on FSDD and Kannada data set. The speech feature vector is calculated using the parameters extracted by Bionic wavelet with different central frequencies of Morlet, Daubechies and Bior3.5, coiflet5 mother wavelets. The obtained Bionic-MFCC optimal features are fed to SVM, ANN and LSTM models for the classification and recognition process. The performance of the models is tabulated for correct recognition that varies from 95% to 96% among these models. The models are tested for various SNRs noise levels like 5 dB, 10 dB, 15 dB and the recognition accuracies of these models are presented for convoluted noisy speech data.

Concept for Mapping Carbon footprint with Change in Vegetation Cover and Population in India

In most of the developing countries, the increasing rate of Carbon emissions is considered as a major cause of concern. India is leading in terms of CO2 emissions as compared to other countries. The vegetation cover comprises only 24.39% of the geographic area of India. Metropolitan cities in India are witnessing rapid urbanization. The primary objective of this proposal is to identify the relationship between the increase in carbon emissions and deforestation in metropolitan cities. An additional objective is to predict the amount of afforestation required for each area to cope up with the carbon emissions over the next 25-years. It can be achieved by using statistical models like ARIMA, LSTM and machine learning techniques such as Random Forest. The proposal provides suggestions on optimal places and techniques for sustainable afforestation to the concerned authorities using artificial intelligence.

Neural network aided fast pointing information determination approach for occultation payloads from in-flight measurements: Algorithm design and assessment

Pointing information is decisive to solving precise profile retrieval issues from occultation measurements. Research regarding stratospheric O3 hole in Antarctic and surface O3 pollution would significantly benefit from massive occultation measurements. A neural network aided pointing information determination approach, in terms of tangent heights, is proposed to address issues requiring fast and easy-to-use determined tangent heights. The geometrical triangular iteration (GTI) algorithm in this work is based on N2 absorption microwindows, and several treatments (e.g., tangential stride generator and triangular-net optimization) are adopted. In addition, LSTM is employed to reduce time consumption and increase accuracy. Atmospheric Chemistry Experiment-Fourier Transform Spectrometer (ACE-FTS) in-flight measurements are used to assess this approach. The comparison between the proposed algorithm and eight official products indicates a promising performance. Correlation coefficient for each orbit is greater than 0.99. The processing time is about 16.6 min per orbit with an average cost of ∼0.06. The introduction of LSTM technique demonstrates an approximate 28.49% better result, with less computation time. It costed less than 30 s to determine eight orbit tangent heights. In general, although minor issues remain, this LSTM-aided GTI algorithm is applicable in industry.