Function irfft 

pub fn irfft<const D: usize>(
    spectrum_re: Tensor<D>,
    spectrum_im: Tensor<D>,
    dim: impl AsIndex,
    n: Option<usize>,
) -> Tensor<D>

Computes the 1-dimensional inverse discrete Fourier Transform for real-valued signals.

This function reconstructs the real-valued time-domain signal from the first non-redundant values ($N/2 + 1$) of the frequency-domain spectrum. Autodiff is supported when the autodiff feature is enabled.

x[n] = (1/N) * Σ X[k] * exp(i*2πkn/N)

Arguments

• spectrum_re - The real part of the spectrum.
• spectrum_im - The imaginary part of the spectrum.
• dim - The dimension along which to take the inverse FFT. Negative dimensions are supported and count from the end.
• n - Optional output signal length. When None, the reconstructed signal length 2 * (size - 1) must be a power of two. When Some(n), n must also be a power of two and the output has exactly n samples. Non-power-of-two n is rejected.

Returns

The reconstructed real-valued signal.

Example

use burn_core::tensor::Tensor;

let device = Default::default();
let real = Tensor::<1>::from_floats([10.0, -2.0, 2.0], &device);
let imag = Tensor::<1>::from_floats([0.0, 2.0, 0.0], &device);
let signal = burn_signal::irfft(real, imag, 0, None);