Internal: Convert FFT frequency bins to Bark/Mel bins
Generate a matrix of weights to combine FFT bins into Bark/Mel bins.
fft2barkmx(nfft, sr = 8000, nfilts = NULL, width = 1, minfreq = 0,
maxfreq = sr/2)
fft2melmx(nfft, sr = 8000, nfilts = 40, width = 1, minfreq = 0,
maxfreq = sr/2, htkmel = FALSE, constamp = FALSE)nfft |
Source FFT size. |
sr |
Sampling rate of the signal. |
nfilts |
Number of desired output frequency bands. If |
width |
Width of each output frequency band in Bark/Mel. |
minfreq |
Minimum frequency. |
maxfreq |
Maximum frequency. |
htkmel |
Use HTK- or Slaney's curve of the Melscale. |
constamp |
Make integration windows peak at 1 ( |
wts |
The weight matrix with 'nfft' columns and 'nfilts' rows. |
binfreqs |
Edge frequencies of the bins. |
While wts has nfft columns, the second half are all zero.
Hence, Bark spectrum is fft2barkmx(nfft, sr) %*% abs(fft(xincols, nfft)).
Sebastian Krey krey@statistik.tu-dortmund.de
Daniel P. W. Ellis: http://www.ee.columbia.edu/~dpwe/resources/matlab/rastamat/, Malcolm Slaney: Auditory Toolbox http://cobweb.ecn.purdue.edu/~malcolm/interval/1998-010/
#Mel matrix in Slaney's mfcc.m: #tuneR:::fft2melmx(512, 8000, 40, 1, 133.33, 6855.5, FALSE, FALSE)
Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.