scholarly journals Keyword Spotting System using Low-complexity Feature Extraction and Quantized LSTM

2021 ◽  
Author(s):  
Kevin Herisse ◽  
Benoit Larras ◽  
Antoine Frappe ◽  
Andreas Kaiser
Keyword(s):  
Feature Extraction ◽  
Low Complexity ◽  
Keyword Spotting

Efficient Domain Search for Fractal Image Compression Using Feature Extraction Technique

2012 ◽  
Vol 488-489 ◽  
pp. 1587-1591
Author(s):  
Amol G. Baviskar ◽  
S. S. Pawale
Keyword(s):  
Feature Extraction ◽  
Image Compression ◽  
Low Complexity ◽  
Contractive Mapping ◽  
Fractal Image Compression ◽  
Compression Technique ◽  
Image Encoding ◽  
Fractal Image ◽  
Coding Method ◽  
The Given

Fractal image compression is a lossy compression technique developed in the early 1990s. It makes use of the local self-similarity property existing in an image and finds a contractive mapping affine transformation (fractal transform) T, such that the fixed point of T is close to the given image in a suitable metric. It has generated much interest due to its promise of high compression ratios with good decompression quality. Image encoding based on fractal block-coding method relies on assumption that image redundancy can be efficiently exploited through block-self transformability. It has shown promise in producing high fidelity, resolution independent images. The low complexity of decoding process also suggested use in real time applications. The high encoding time, in combination with patents on technology have unfortunately discouraged results. In this paper, we have proposed efficient domain search technique using feature extraction for the encoding of fractal image which reduces encoding-decoding time and proposed technique improves quality of compressed image.

A unified framework for the design of low-complexity wavelet filters

2017 ◽  
Vol 15 (06) ◽  
pp. 1750054 ◽  
Author(s):  
Ameya K. Naik ◽  
Raghunath S. Holambe
Keyword(s):  
Feature Extraction ◽  
Compact Support ◽  
Low Complexity ◽  
Design Parameters ◽  
Unified Framework ◽  
Vanishing Moments ◽  
Wavelet Filters ◽  
Design Variables ◽  
Optimum Filter ◽  
Simulation Results

An outline is presented for construction of wavelet filters with compact support. Our approach does not require any extensive simulations for obtaining the values of design variables like other methods. A unified framework is proposed for designing halfband polynomials with varying vanishing moments. Optimum filter pairs can then be generated by factorization of the halfband polynomial. Although these optimum wavelets have characteristics close to that of CDF 9/7 (Cohen-Daubechies-Feauveau), a compact support may not be guaranteed. Subsequently, we show that by proper choice of design parameters finite wordlength wavelet construction can be achieved. These hardware friendly wavelets are analyzed for their possible applications in image compression and feature extraction. Simulation results show that the designed wavelets give better performances as compared to standard wavelets. Moreover, the designed wavelets can be implemented with significantly reduced hardware as compared to the existing wavelets.

scholarly journals A Comparison of Low-Complexity Real-Time Feature Extraction for Neuromorphic Speech Recognition

2018 ◽  
Vol 12 ◽  
Author(s):  
Jyotibdha Acharya ◽  
Aakash Patil ◽  
Xiaoya Li ◽  
Yi Chen ◽  
Shih-Chii Liu ◽  
...  
Keyword(s):  
Feature Extraction ◽  
Speech Recognition ◽  
Real Time ◽  
Low Complexity

scholarly journals Vehicle Detection In Remote Sensing Images

2019 ◽  
Vol 8 (11) ◽  
pp. 928-933
Keyword(s):  
Feature Extraction ◽  
Vehicle Detection ◽  
Low Complexity ◽  
Traffic Monitoring ◽  
Color Spaces ◽  
Continuous Increase ◽  
The Road ◽  
Uav Images ◽  
Candidate Regions ◽  
Detection And Localization

Traffic monitoring and management is one of the most crucial tasks of governing bodies in modern big cities. With each passing day the traffic problem grows in complexity due to the continuous increase of participating vehicles and the hard expansion of the road network and parking places. In this article we introduce a new method for vehicle detection and localization in parking lots using high resolution UAV images. In order to end up with practical and yet effective approach, which could be implemented on low computing hardware resources and integrated with the camera in the UAV, we considered simple steps in the proposed algorithm for optimization. It follows the machine learning pipeline such as preprocessing, sensing, feature extraction, training and classification. In preprocessing the images are thresholded iteratively in multiple color spaces to extract the candidate regions of interest (ROI). The algorithm relies on point and shape features using fast techniques in the feature extraction. The features are then clustered by the K-means algorithm and represented by the resulted clusters’ centers. Region based linear classification is finally applied using SVM to classify if the object is a vehicle or else. The proposed approach proved high detection and classification accuracy more than 86% and still running under the low complexity constraint..

Circulant Matrix-Based Continuous Wavelet Transform for Achieving Low Complexity Electrocardiogram Feature Extraction in Health Monitoring Applications

2020 ◽  
Vol 17 (1) ◽  
pp. 254-259
Author(s):  
Harikrishna Ponnam ◽  
Jakeer Hussain Shaik
Keyword(s):  
Feature Extraction ◽  
Wavelet Transform ◽  
Computational Complexity ◽  
Error Rate ◽  
Continuous Wavelet Transform ◽  
Low Complexity ◽  
Circulant Matrix ◽  
Continuous Wavelet ◽  
Ecg Signals ◽  
Gradient Based

In the application of remote cardiovascular monitoring, the computational complexity and power consumption need to be maintained in a considerable level in order to prevent the limitations introduced by the computationally constrained equipment’s that perform the process of continuous monitoring and analysis. In this paper, a Circulant Matrix-based Continuous Wavelet Transform (CM-CWT)-based feature extraction mechanism is contributed to minimizing the computational complexity incurred during the process of feature extraction from the input ECG signals. This proposed CM-CWT mechanism derives the advantages of the Circulant Matrix-based Continuous Wavelet Transform and Gradient-based filtering design for achieving excellent feature extraction from ECG signals with low computational complexity. The experimental investigation of the proposed CM-CWT mechanism is conducted using the factors of computational complexity, sensitivity, prediction accuracy and error rate for estimating its predominance over the compared DWT-HAAR and HIFEA approaches used for ECG feature extraction. The experiments of the proposed CM-CWT mechanism on an average is estimated to reduce the error rate to the maximum of 21% compared to the existing DWT-HAAR and HIFEA approaches used for ECG feature extraction.

scholarly journals Low-Power Audio Keyword Spotting Using Tsetlin Machines

2021 ◽  
Vol 11 (2) ◽  
pp. 18
Author(s):  
Jie Lei ◽  
Tousif Rahman ◽  
Rishad Shafik ◽  
Adrian Wheeldon ◽  
Alex Yakovlev ◽  
...  
Keyword(s):  
Low Power ◽  
Learning Algorithm ◽  
Learning Automata ◽  
Low Complexity ◽  
Keyword Spotting ◽  
High Learning ◽  
Memory Footprint ◽  
Learning Efficacy ◽  
On Chip ◽  
Machine Interaction

The emergence of artificial intelligence (AI) driven keyword spotting (KWS) technologies has revolutionized human to machine interaction. Yet, the challenge of end-to-end energy efficiency, memory footprint and system complexity of current neural network (NN) powered AI-KWS pipelines has remained ever present. This paper evaluates KWS utilizing a learning automata powered machine learning algorithm called the Tsetlin Machine (TM). Through significant reduction in parameter requirements and choosing logic over arithmetic-based processing, the TM offers new opportunities for low-power KWS while maintaining high learning efficacy. In this paper, we explore a TM-based keyword spotting (KWS) pipeline to demonstrate low complexity with faster rate of convergence compared to NNs. Further, we investigate the scalability with increasing keywords and explore the potential for enabling low-power on-chip KWS.

scholarly journals A MULTI THREADED FEATURE EXTRACTION TOOL FOR SAR IMAGES USING OPEN SOURCE SOFTWARE LIBRARIES

2018 ◽  
Vol XLII-5 ◽  
pp. 155-160
Author(s):  
C. Bipin ◽  
C. V. Rao ◽  
P. V. Sridevi ◽  
S. Jayabharathi ◽  
B. G. Krishna
Keyword(s):  
Remote Sensing ◽  
Feature Extraction ◽  
Computational Complexity ◽  
Remote Sensing Data ◽  
Low Complexity ◽  
Higher Order ◽  
Sar Images ◽  
Automatic Feature Extraction ◽  
Software Libraries ◽  
Computationally Intensive

<p><strong>Abstract.</strong> In this paper, we propose a software architecture for a feature extraction tool which is suitable for automatic extraction of sparse features from large remote sensing data capable of using higher order algorithms (computational complexity greater than <i>O</i>(<i>n</i>)). Many features like roads, water bodies, buildings etc in remote-sensing data are sparse in nature. Remote-sensing deals with a large volume of data usually not manageable fully in the primary memory of typical workstations. For these reason algorithms with higher computational complexity is not used for feature extraction from remote sensing images. A good number of remote sensing applications algorithms are based on formulating a representative index typically using a kernel function which is having linear or less computational complexity (less than or equal to <i>O</i>(<i>n</i>)). This approach makes it possible to complete the operation in deterministic time and memory.</p><p>Feature extraction from Synthetic Aparture Radar (SAR) images requires more computationally intensive algorithm due to less spectral information and high noise. Higher Order algorithms like Fast Fourier Transform (FFT), Gray Level Co-Occurrence Matrix (GLCM), wavelet, curvelet etc based algorithms are not preferred in automatic feature extraction from remote sensing images due to their higher order of computational complexity. They are often used in small subsets or in association with a database where location and maximum extent of the features are stored beforehand. In this case, only characterization of the feature is carried out in the data.</p><p>In this paper, we demonstrate a system architecture that can overcome the shortcomings of both these approaches in a multi-threaded platform. The feature extraction problem is divided into a low complexity with less accuracy followed by a computationally complex algorithm in an augmented space. The sparse nature of features gives the flexibility to evaluate features in Region Of Interest (ROI)s. Each operation is carried out in multiple threads to minimize the latency of the algorithm. The computationally intensive algorithm evaluates on a ROI provided by the low complexity operation. The system also decouples complex operations using multi-threading.</p><p>The system is a customized solution developed completely in python using different open source software libraries. This approach has made it possible to carry out automatic feature extraction from Large SAR data. The architecture was tested and found giving promising results for extraction of inland water layers and dark features in ocean surface from SAR data.</p>

Sign in / Sign up

Export Citation Format

Share Document