Simulated Random Level Stepwise Signal Dataset

This dataset provides simulated stepwise signals characterized by random transition amplitudes and dynamic dwell times. The signals are produced using a continuous-time generation logic, where the state dwell times consist of an exponentially distributed random waiting time combined with a fixed minimum duration constraint, and the signal jumps to a completely random continuous value uniformly distributed between 0 and 1.

traindata– The dataset for training model development, containing signals with SNR levels of 3, 5, and 7.

testdata– The dataset for evaluation and testing, containing signals with SNR levels of 1, 3, and 5.

File Naming Convention

The specific parameters used to generate each set of signals are indicated directly within the filenames:

snr: The Signal-to-Noise Ratio is specified in the filename to assess model performance under different noise conditions.

pro_: Indicates the base transition probability used during signal generation. This value dictates the average number of jumps per second, directly controlling the temporal frequency of the step transitions.

Signal Properties

Duration & Sampling: All trajectories have a total duration of 1000 seconds and were sampled at 1 Hz (totaling 1000 data points per trace).

Amplitude Distribution: Unlike discrete-state models, the amplitude of every new state is drawn randomly from a continuous uniform distribution [0, 1].

Dwell Time Constraint: To simulate realistic physical transitions, the total duration of any given state is constrained by a minimum hold time of 2.0 seconds, combined with a random exponential waiting time derived from the transition probability.