mel: Convert Hz to the mel scale

In emuR: Main Package of the EMU Speech Database Management System

Convert Hz to the mel scale

Description

The calculation is done using the formulae mel = 1/log(2) * (log(1 + (Hz/1000))) * 1000 where Hz is the frequency in Hz.

Usage

mel(a)

Arguments

a	A vector or matrix of data or a spectral object.

Details

If 'data' is a spectral object, then the frequencies are changed so that they are proportional to the mel scale and such that the mel intervals between frequencies are constant between the lowest and highest frequencies. More specifically, suppose that a spectral object has frequencies at 0, 1000, 2000, 3000, 4000 Hz. Then the corresponding frequencies extend in mel between 0 and 2321.928 mel (=4000 Hz in mels) in four equal intervals, and linear interpolation is used with the 'approx' function to obtain the dB values at those frequencies.

Value

A vector or matrix or spectral object of the same length and dimensions as data.

Author(s)

Jonathan Harrington

References

Traunmueller, H. (1990) "Analytical expressions for the tonotopic sensory scale" J. Acoust. Soc. Am. 88: 97-100.

See Also

bark, plot.spectral

Examples

 #convert Hertz values to mel

 vec <- c(500, 1500, 2500)
 vec
 mel(vec)

        
        # convert Hertz values to mel 

        mel(vec)
        

        # convert the $data values in a trackdata object to mel
        # create a new track data object

        t1 <- dip.fdat
        t1[1]
        
        # convert Hertz to mel

        t1$data <- mel(t1$data)
        t1[1]

# warp the frequency axis of a spectral object such
# that it is proportional to the mel scale.

w = mel(e.dft)
oldpar = par(mfrow=c(1,2))
plot(w, type="l")


# The values of w are at equal mel intervals. Compare
# with

plot(e.dft, freq=mel(trackfreq(e.dft)))

# the latter has a greater concentration of values
# in a higher frequency range.

par(oldpar)

emuR documentation built on Nov. 4, 2023, 1:06 a.m.