This work is focused on single word Arabic automatic speech recognition (AASR). Two techniques are used during the feature extraction phase; Log frequency spectral coefficients (MFSC) and Gammatone-frequency cepstral coefficients (GFCC) with their first and second-order derivatives. The convolutional neural network (CNN) is mainly used to execute feature learning and classification process. CNN achieved performance enhancement in automatic speech recognition (ASR). Local connectivity, weight sharing, and pooling are the crucial properties of CNNs that have the potential to improve ASR. We tested the CNN model using an Arabic speech corpus of isolated words. The used corpus is synthetically augmented by applying different transformations such as changing the pitch, the speed, the dynamic range, adding noise, and forward and backward shift in time. It was found that the maximum accuracy obtained when using GFCC with CNN is 99.77 %. The outcome results of this work are compared to previous reports and indicate that CNN achieved better performance in AASR.
This paper presents a process for cursive Arabic writing recognition. This problem is complicated by the wide variety of styles, personal attitudes, and multi-level writing. Therefore, we implement a hierarchical strategy for recognition. Simultaneously, this requires the definition of primitives, connectives, and laws of concatenation. Consequently, we develop a linguistic recognition system together with its grammar and parser.The presented system, a new approach to this important area of research that reveals the complexities and advantages of Arabic cursive writing, then providing adequate solutions to the encountered difficulties.The resulting pattern grammar belongs to the context-free class. We complement the process by a lexicon, whose entries are the chosen primitives. The resulting system has a corresponding non-deterministic recursive automaton implementation.
The success of model based object recognition in a closed world depends on the correct choice of parametric primitive shapes or correct choice of a set of basic features. The later could be the result of machine learning. In this paper a choice of parametric primitive shapes is made and a recognition procedure including a set of developed algorithms is presented. The procedure is in essence a divide and conquer paradigm. The scene is divided into a two sets: (a) set of flat surfaces, and (b) set of objects. The point clouds corresponding to individual objects are extracted. Each extracted cluster is further analyzed into: (a) candidate parametric object, (b) undetermined object parts such as handles, and (c) noise due to sensory and registration errors that represent high frequency signal to be filtered out. This procedure is in the same time an autonomous learning paradigm that enables to memorize the recognized objects and uses this information to answer future questions about the scene. The experimental part is introduced to verify the proposed procedure.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.