Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

need help in classification.....

Status
Not open for further replies.

vjfaisal

Full Member level 4
Joined
Sep 24, 2006
Messages
205
Helped
0
Reputation
0
Reaction score
0
Trophy points
1,296
Location
pakistan
Activity points
2,701
hi

please help in choosing which method is most efficient for the clasification of EEG signal.

1) neural network

2) ICA

Is there any orther technique , please mnetion those......

best regards
 

Well, the amount of techniques is really enormous. Everything depends on the concrete goal you want to achieve with your classification. I'll mention several groups of algorithms and afterwards you'll choose the necessary ones according to your task. In general, all these methods refer to the so-called Data Mining, which tries to discover new, non-trivial, practically useful and adjusted for human's interpretation knoledge just from initial data sets.

1) Segmentation algorithms. Segmentation divides the signal (the data set) in to a number of fragments with different length, each of which possesses the uniform characteristics. Usually segmentation is applied during the step of preliminary exploration of signals in order to find out the boundaries of sements. It's considered that different segments correspond to various types of processes (noise, trend, transmission, etc.)

Segmentation algorithms: 1) Autoregressive models
2) Principal component segmentation
3) Statistical segmentation (AIC,BIC, MDL criterions, etc).
4) Cumulative sum segmentation
5) Segmentation based on smoothing

and others.

2) Clustering. Cluster-analysis is used in order to find out the segments , which are turned out to refer to the same classes, hence, they posses the same or very close properties (statistical, spectral, etc.). Clustering is used after segmentation. The majority of algorithms is based either on computing the centres of clusterers and combining the vectors, which represent the same class, or on automatical clusterisation, when the best and most suitable number of clusterers is established adaptively. The most popular algorithms are:

1) k-means, k-median
2) EM
3) FartherstFirst
4) CobWeb
5) Hierarchical clustering
6) Statistical clustering
7) Graph models

3) Classification

Imagine you've performed clustering. Then you get a new fragment (or even a new sample) and you need to find out, which class it belongs to. This problem is solved by powerful algorithms of classification. The results may be expressed in rules and trees. The algorithms:

1) J48
2) Naive Bayes

4) Sequential analysis

This direction in Data Mining tries to sort all events according to their time arisal. This denotes, that it searches the so-called time-patterns, which are existed in 13 forms (according to Allen's logics). Sequential analysis helps answer the question which patterns are earlier or later than the others

_______________________________________________

I really advise you to get acquanted or to read any book about Data Mining

With respect,

Dmitrij
 

hi

i want to do classfication in EEG data , My project is of Brain Computer Interface.

So, plz guide me which algorithm or techniques you memntioned have a best results in this case , OR kindly mention which could be best one in BRAIN COMPUET INTERFACE.




best regards
 

Unfortunately, I dont't study EEG signal processing. I'm interested in exploring telemetric signals, therefore it's really difficult for me to give use a suitable advice. Nevertheless, you can carefully try to apply all the mentioned algorithms (of course, it'll take much time, but this time sacrifice will be refunded subsequently) to the model signals and estimate the final results. If you're satisfied with them, the you may use this algorithms afterwards. Otherwise, the algorithms will be admitted to be useless.

With respect,

Dmitrij
 

aNYONE WHO KNOWS THIS PLEASE HELP ME............



REGARDS
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top