Problems in Aki Härmä's publications! (be aware of this guy)

This is certainly not a complete errata but anyway an attempt to do something before the same errors propagate to other people's articles.

IEEE Trans. SAP articles (July, 2001) and (Nov, 2001)

Prof. W. Bastiaan Kleijn kindly helped me to understand that there is an error in the spectrum interpretation of WLP and generalized LPC versions in some of my articles. In (July, 2001) the problem is in Eq. (13) (d\tilde{omega} should be d\omega in the second expression). In (Nov, 2001) the same problem starts with Eq. (28) and is probably infiltrated all the way to (30-31). Note that the same error is can also be found in some other authors' (older) articles. Fortunately, the problem is restricted to the continuous-time representation with integrals and is not leaking to the z domain. A detailed errata will be available here shortly.

ICASSP 1997

In my ICASSP 1997 article, Figure 4 a) shows a closed-loop CWLPC encoder while an open-loop system was actually used. This is a ridiculous error and I can't recall how it was produced. The text is clean.

Signal Processing, 2000

Alexis Glass from the Kyushu Institute of Design, Japan, kindly advised that there may be a bug in algorithmic descriptions of the direct implementation of the warped IIR lattice filter introduced in my (Signal Processing, 2000) article. I have not checked this but the fix he provided is below:


1) In figure 3a, the signal from 'b1' should be subtracted from 'f0', and
likewise the signal from 'f0' path should be subtracted from 'b1'.  (And so
on, down the filter)


 
2) Subsequently, in algorithm 3:

Si = v*r[i] - (S[i-1]-k[i-1]*p[i-2])


3) Also as a result of the changes in 1), algorithm 4 should be written as
follows:

x=x+k[i]*s;
b=s-k[i]*f;
f=f-k[i]*s;

Back to the publications page .

Aki Harma

Last modified: Mon Nov 18 11:54:26 EET 2002