Skip to content

Abstract Binned Stimulus Generation Method

Abstract class describing all features common to a stimulus generation method that uses a binned representation of the signal.

Abstract Properties

These properties are automatically instantiated for subclasses, since they are not abstract themselves. Default values are given:

- n_bins = 100 % The number of bins to break the frequency spectrum into 
- unfilled_dB = -100 % The dB value of "unfilled" bins
- filled_dB = 0 % The dB value of "filled" bins

get_freq_bins

[binnum, Fs, nfft, frequency_vector] = self.get_freq_bins()

Generates a vector indicating which frequencies belong to the same bin, following a tonotopic map of audible frequency perception.

OUTPUTS:

  • binnum: n x 1 numerical vector that contains the mapping from frequency to bin number e.g., [1, 1, 2, 2, 2, 3, 3, 3, 3, ...]

  • Fs: 1x1 numerical scalar, the sampling rate in Hz

  • nfft: 1x1 numerical scalar, the number of points of the full FFT

  • frequency_vector: n x 1 numerical vector. the frequencies that binnum maps to bin numbers

See Also

get_empty_spectrum

[spect] = self.get_empty_spectrum()

OUTPUTS:

  • spect: n x 1 numerical vector, where n is equal to the number of fft points (nfft) and all values are set to unfilled_dB.

See Also

spect2binnedrepr

Get the binned representation which is a vector containing the amplitude of the spectrum in each frequency bin.

ARGUMENTS:

  • T: n_frequencies x n_trials matrix representing the stimulus spectra

OUTPUTS:

  • binned_repr: n_trials x n_bins matrix representing the amplitude for each frequency bin for each trial

See Also

binnedrepr2spect

Get the stimuli spectra from a binned representation.

ARGUMENTS:

  • binned_repr: n_bins x n_trials representing the amplitude in each frequency bin for each trial

OUTPUTS:

  • T: n_frequencies x n_trials representing the stimulus spectra

See Also

binnedrepr2wav

Get the peak-sharpened waveform of a binned representation. Can pass n sounds and the modifications will be applied to all.

ARGUMENTS:

  • binned_repr: n_bins x n numerical vector representing the amplitude in each frequency bin.
  • mult: n x 1 vector or scalar, the peak sharpening factor corresponding to each binned_repr.
  • binrange: n x 1 vector or scalar, must be between [1, 100], the upper bound of the dynamic range of the stimuli from [0, binrange] corresponding to each binned_repr.
  • new_n_bins: 1 x 1 scalar, default: 256, the number of bins to upsample to before synthesis.
  • filter: logical, name-value, default: false, flag to filter the waveform.
  • cutoff: n x 2, name-value, default: [2000 self.max_freq], min and max cutoff frequencies corresponding to each binned_repr. If values satisfy min > 0 && max < self.max_freq, bandpass filter is used. If only min < 0, highpass is used. Otherwise, lowpass.
  • order: 1 x 1 positive integer, name-value, default: 5, the filter order.

OUTPUTS:

  • wav: nfft+1 x 1 numerical vector representing the upsampled, peak-sharpened wavform of the binned representation.
  • X: nfft/2 x 1 numerical vector, the upsampled, peak-sparpened spectrum of the binned representation.

See Also

bin_signal

W = self.bin_signal(W)

Inputs a waveform, converts to a spectrum, bins the spectrum, and then converts back to a waveform.

ARGUMENTS:

W: n x 1 numerical vector, the waveform Fs: 1x1 numerical scalar, the sample rate

See Also