applications of artificial neural networks in voice recognition and nettalk
DESCRIPTION
Applications of Artificial Neural Networks in Voice Recognition and Nettalk..detailed description of artificial neuar networkTRANSCRIPT
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
APPLICATIONS OF ARTIFICIAL NEURALNETWORKSAPPLICATIONS
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Artificial Neural NetworkWhy Use Neural Networks?Advantages
What is a Neural Network?
A neural network is a processing device, either analgorithm, or actual hardware, whose design wasmotivated by the design and functioning of human brainsand components thereof.The abilities of different networks can be related to theirstructure, dynamics and learning methods.There are many different types of Neural Networks, eachof which has different strengths particular to theirapplications.
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Artificial Neural NetworkWhy Use Neural Networks?Advantages
Artificial Neural Network
Historical background.First Attempts.Promising and Emerging TechnologyToday
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Artificial Neural NetworkWhy Use Neural Networks?Advantages
Features Of ANN
Their ability to represent nonlinear relations makes themwell suited for non linear modeling in control systems.Adaptation and learning in uncertain system through offline and on line weight adaptation Parallel processingarchitecture allows fast processing for large-scale dynamicsystem.Neural network can handle large number of inputs and canhave many outputs.
SHAJEER.K.B ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Artificial Neural NetworkWhy Use Neural Networks?Advantages
Advantages
Adaptive learningSelf-OrganisationReal Time OperationFault Tolerance via Redundant Information Coding
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
From Human Neurons to Artificial NeuronsNeural Network ModelsNeural Network Models
From Human Neurons to Artificial Neurons
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
From Human Neurons to Artificial NeuronsNeural Network ModelsNeural Network Models
From Human Neurons to Artificial Neurons
InputsWeightsThresholdActivation Function
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
From Human Neurons to Artificial NeuronsNeural Network ModelsNeural Network Models
Multilayer perceptron
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
From Human Neurons to Artificial NeuronsNeural Network ModelsNeural Network Models
Two hidden layer multilayer perceptron
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
Applications
Voice Recognition - Transcribing spoken words into ASCIItext.Image Compression - Because neural networks canaccept a vast array of input at once, and process it quickly,they are useful in image compression.Medical Diagnosis - Assisting doctors with their diagnosisby analyzing the reported symptoms and/or image datasuch as MRIs or X-rays.
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
Process Modeling and Control - Creating a neural networkmodel for a physical plant then using that model todetermine the best control settings for the plantTargeted Marketing - Finding the set of demographics,which have the highest response rate for a particularmarketing campaign.Financial forecasting - Using the historical data of asecurity to predict the future movement of that security.
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
Voice Recognition
Speech recognition: The objective is to determine thesequence of sound units from the speech signal so that thelinguistic message in the form of text can be decoded fromthe speech signal.Speech synthesis: The objective is to determine thesequence of sound units corresponding to a text so that agiven text message can be encoded to a speech signal.Speaker identification: the objective is to determine theidentity of the speaker from the speech signal.
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
NETTALK
used to generate synthetic speechdeveloped by Terry Sejnowski and Charles RosenbergThe data ârepresentation scheme employed allows atemporal pattern sequence to be represented spatiallyNETtalk Data Representation
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
NETTALK
ConsiderExample:FIND FIEND FRIEND FEINT
a sliding window technique for representing words aspatterns
SHAJEER.K.B ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
Neural networks:medicine
SHAJEER.K.B ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
Neural networks in medicine
Artificial Neural Networks (ANN) are currently a ’hot’research area in medicine and it is believed that they willreceive extensive application to biomedical systems in thenext few years.At the moment, the research is mostly on modelling partsof the human body and recognising diseases from variousscans (e.g. cardiograms, CAT scans etc.)Neural networks are ideal in recognising diseases usingscans since there is no need to provide a specificalgorithm on how to identify the disease.
SHAJEER.K.B ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
CARDIOVASCULAR SYSTEM:MODELLING
Neural Networks are used experimentally to model thehuman cardiovascular system.Diagnosis can be achieved by building a model of thecardiovascular system of an individual and comparing itwith the real time physiological measurements taken fromthe patient.A model of an individual’s cardiovascular system mustmimic the relationship among physiological variables (i.e.,heart rate,blood pressures, and breathing rate) at differentphysical activity levels.If a model is adapted to an individual, then it becomes amodel of the physical condition of that individual.
SHAJEER.K.B ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
ELECTRONIC NOSES:DIAGNOSING
Electronic noses have several potential applications intelemedicine.The electronic nose would identify odours in the remotesurgical environment.Telemedicine is the practice of medicine over longdistances via a communication link.
SHAJEER.K.B ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
Image Compression
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
BOTTLENECK ARCHITECTURE
SHAJEER.K.B ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
QUANTIZATION
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
HOW DOES A NETWORK LEARN TO DO THIS?
The goal of these data compression networks is tore-create the input itself.This data is presented over and over, and the weightsadjusted, until the network reproduces the image relativelyfaithfully.Once training is complete, image re-construction isdemonstrated in the recall phase.we can continue to train the network if the output is not ofhigh enough quality.
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
ORIGINAL IMAGERECONSTRUCTED
IMAGE
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
PROCESS MODELLING AND CONTROLAPPLICATIONDS
ANN is able to emulate the information processingcapabilities of biological neural system.Artificial neural networks are implemented as softwarepackages in computers and being used to incorporate ofartificial intelligence in control system.ANN has overcome many of the difficulties that tconventional adaptive control systems suffer while dealingwith non linear behavior of process.
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
ANN DESIGN PARAMETERS INCLUDE
The interconnection strategy/network topology/networkstructure.Unit characteristics (may vary within the network andwithin subdivisions within the network such as layers).Training procedures.Training and test sets.Input/output representation and pre- and post-processing.
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
LEARNING TECHNIQUES
Forward modelingThe basic configuration used for non-linear systemmodeling and identification using neural network.The number of input nodes specifies the dimensions of thenetwork input.In system identification context, the assignment of networkinput and output to network input vector.
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
DIRECT INVERSE MODELING
This approach employs a generalized model suggested byPsalters et al.to learn the inverse dynamic model of theplant as a feed forward controller.Here, during the training stage, the control input arechosen randomly within there working range.The corresponding plant output values are stored, as atraining of the controller cannot guarantee the inclusion ofall possible situations that may occur in future.
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
Targeted Marketing and Financial Forecasting
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
TARGETED MARKETING AND FINANCIALFORECASTING
A simple back-propagation network of three layers, and it istrained and tested on a high volume of historical marketdata.The challenge here is not in the network architecture itself,but instead in the choice of variables and the informationused for training. Neural Networks in businessUsing neural networks for business purposes, includingresource allocation and scheduling.There is also a strong potential for using neural networksfor database mining, that is, searching for patterns implicitwithin the explicitly stored information in databases.
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
Applications in SpeechNeural Networks in MedicineImage CompressionProcess Modelling and ControlTargeted Marketing and Financial Forecasting
MARKETING
The Airline Marketing Tactician is a computer system madeof various intelligent technologies including expertsystems.A feed forward neural network is integrated with the AMTand was trained using back-propagation to assist themarketing control of airline seat allocations.The adaptive neural approach was amenable to ruleexpression.The application’s environment changed rapidly andconstantly, which required a continuously adaptive solution.
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
ReferenceGratitude
CONCLUSION
The computing world has a lot to gain from neuralnetworks.The neural network can be used to design a controller,robotics, industrial manufacturing, aerospace and severalothers.The ability of a feedback network to store patterns can beimproved, if we can exploit the chaotic nature of thenetwork dynamics.Due to inherent non-linearity and also due to the learningability, neural networks appear to be promising in somedecision making applications.
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
ReferenceGratitude
REFERENCE
Artificial Neural Networks and its Applications Girish KumarJha I.A.R.I Newdelhi 110 012 [email protected] Networks:A Comprehensive Study 2nd EditionSimon HaykinArtificial Neural Networks:B Yegananarayana
ARTIFICIAL NEURAL NETWORKS
INTRODUCTIONFROM HUMAN NEURONS TO ARTIFICIAL NEURONS
APPLICATION FIELDSCONCLUSION
ReferenceGratitude
Thank You
ARTIFICIAL NEURAL NETWORKS