next Uses of Linear Predictive Coding (LPC)
up Music 270a: Signal Analysis
previous Pitch Detection


Harmonic Product Spectrum (HPS)

HPS (Noll 1969) measures the maximum coincidence for harmonics for each spectral frame according to

$\displaystyle Y(\omega) = \prod^R_{r=1} \vert X(\omega r)\vert,$ (1)

where R is the number of harmonics being considered.

The resulting periodic correlation array $ Y(\omega)$ is then searched for a maximum value of a range of possible fundamental frequencies $ \omega_i$

$\displaystyle \hat{Y} = \max_{\omega_i}{Y(\omega_i)}$ (2)

to obtain the fundamental frequency estimate.

Octave errors are common (detection is sometimes an octave too high).

To correct, apply this rule: if the second peak amplitude below initially chosent pitch is approximately 1/2 of the chosen pitch AND the ratio of amplitudes is above a threshold (e.g., 0.2 for 5 harmonics), THEN select the lower octave peak as the pitch for the current frame.

Due to noise, frequencies below about 50 Hz should not be searched for a pitch.

Pros: HPS is simple to implement, does well under a wide range of conditions, and runs in real-time.

Cons: low frequency resolution must be enhanced by zero-padding, so that the spectrum can be interpolated to the nearest semitone. This means that high frequencies are also being unecessarily interpolated.

See hps.m

Figure 2: Harmonic Product Spectrum
\scalebox{0.8}{\includegraphics{eps/hps1.eps}}


next Uses of Linear Predictive Coding (LPC)
up Music 270a: Signal Analysis
previous Pitch Detection

``Music 270a: Signal Analysis'' by Tamara Smyth, Department of Music, University of California, San Diego.
Download PDF version (analysis.pdf)
Download compressed PostScript version (analysis.ps.gz)
Download PDF `4 up' version (analysis_4up.pdf)
Download compressed PostScrip `4 up' version (analysis_4up.ps.gz)

Copyright © 2015-11-23 by Tamara Smyth.
Please email errata, comments, and suggestions to Tamara Smyth<trsmyth@ucsd.edu>