Training a neural network for time series data classification

I have a dataset acquired through 8 sensors. I record the data for a certain period of time from 8 sensors simultaneously. I have repeated this process for 30 times. So in all i have 30 sets of time series data from 8 sensors. I want to train a neural network and classify the time series qualitatively as 'good' or 'bad' i.e. 0 or 1.

Till the best of my knowledge I should use TDNN architecture. Can you point me to any other architectures that might be suitable for this task?

I tried working with ntstool. It throws up error and the reason is data structure issues with input data and target data. Can anyone explain me the expected data structure and size of target data and input data for my case? (i know it accepts cell {}, but conceptual knowledge in simple words with numbers mentioned above would be helpful)

I am also unclear as to how TDNN works. Any pointers/simple explaination to as how it accepts inputs and the training process would be more than helpful.

[abdenour]

hi, if the inputs of your data set is fixed size (the number of time steps can be assumed to be vector size) and normalized, meaning that test and training data are in the same scale and centered around a given mean, you can just apply pattern recognition tool in matlab neural net toolbox. Otherwise, you should look for more involving neural nets architecture for time series classification such as LSTMs. you may find this link useful https://github.com/huashiyiqike/LSTM-MATLAB

good luck,