Why is the accuracy on the test dataset very low when training a neural network on an IMU dataset?

I am trying to train an IMU (Inertial Measurement Unit) dataset. The dataset contain 6 features (3-gyro, 3-accelerometer) and 1 label column. I have build a neural network via Conv1D, LSTM and Dense layers respectively (in Keras Framework). The train set accuracy achieve %89 accuracy but validation accuracy is stucked on %27. I know it means overfitting. I tried to use dropout layers and hyperparamter tunning but validation accuracy is still low. The model can’t predict good solutions if I test on untrained data(test set). What can I do to increase accuracy?

You can check my notebook and dataset here.

Will the accuracy on the test dataset increase if we add noise on the tabular data?

Or, do I need to have more data to make a generalization and have good test accuracy?