Channel file format specifications

UWA-Channels MAT File Format Specification

The uwa-channels-compatible .mat files must be saved with the following flags:

  • -v7.3: Supports large variables and the HDF5 file format, ensuring better compatibility with other programming languages such as Python and Julia.
  • -nocompression: Accelerates loading speed.

Required Fields

Each .mat file must include the following variables:

h_hat

  • Type: Multi-dimensional numeric tensor (single or double precision). Arbitrarily scaled.
  • Dimensions: [delay, receiver, time]
  • Description: The estimated time-varying channel impulse response (TVIR), complex baseband.

params

A MATLAB structure containing the following scalar fields:

  • fs_delay (Hz): Sampling rate of h_hat along the delay axis.
  • fs_time (Hz): Sampling rate of h_hat along the time axis.
  • fc (Hz): The probing waveform was basebanded with respect to fc prior to the TVIR estimation; fc typically corresponds to the center frequency of the probing waveform.

version

  • Type: Numeric scalar
  • Description: Dataset format version number.

Optional Fields

theta_hat

  • Type: Numeric matrix of size [receiver, time] (single or double precision).
  • Sampling Rate: Must match params.fs_delay
  • Duration Constraint: The time duration of theta_hat must match the third dimension of h_hat, i.e., size(theta_hat, 2) * params.fs_time == size(h_hat, 3) * params.fs_delay.
  • Description: For each hydrophone (receiver), theta_hat represents a time-varying resampling factor $R(t)$ to be applied to the output signal in conjunction with $h(t,\tau)$: $$ R(t) = 1 - \frac{1}{2 \pi f_c} \frac{\text{d} \hat{\theta}(t)}{\text{d} t} $$ where $f_c$ is the center frequency, and the $\hat{\theta}(t)$ is the theta_hat.

f_resamp

  • Type: Scalar (single or double precision)
  • Units: Unitless resampling factor
  • Description: A time-invariant resampling factor applied to the output signal after convolution. This is typically the inverse of a resampling operation used to remove nominal Doppler offsets during channel estimation. It is applied after the time-varying convolution.

meta

The meta structure is optional but encouraged for documenting dataset context. The suggested fields include:

  • meta.experiment_name (string): Name of the experiment.
  • meta.experiment_info (string or struct): Descriptive details about the experimental setup.
  • meta.array_info (string or struct): Information about the hydrophone array (e.g., geometry, number of elements).
  • meta.authorship (string or struct): Data contributor names and affiliations.

Users are free to add any number of additional fields to meta to record experiment-specific metadata.

Dataset Configurations

The following configurations are currently supported:

  • h_hat only — supported but not used in official datasets.
  • h_hat and theta_hat — used in the blue, red, green, purple, yellow, abyssal, and namikaze datasets.
  • h_hat and f_resamp — used in Watermark datasets.
  • h_hat, theta_hat, and f_resamp — used in the Sky dataset.