Subject: Re: musical tones in speech From: "Alain de Cheveigne'" <Alain.de.Cheveigne(at)IRCAM.FR> Date: Fri, 11 May 2001 18:23:58 +0200Bruno, I didn't miss no boat. It's simply that, having spent many a day visually checking the F0 data, I'm dammed if I'm going to go back through them manually assigning targets. Either my test is relevant (and negative), or someone must explain how targets are distributed according to the musical scale without there being any evidence of this structure in the actual data that the targets are derived from. As to how such an artifact could arise, a possibility is that the software that was used to choose targets quantized F0 values to semitones. The fact that the distribution (Braun's fig. 2) did not consist of spikes at semitones would be explained by the fact that the histograms are of "mean-speech target frequencies", these means being taken over "16 readings of different samples". This is a guess, the paper is unclear on how all these things were done. Alain --- >The interchange between Martin and Alain is most interesting. I have >the feeling that Alain has missed the boat by looking at complete F0 >contours, but his comments do raise the following question: Is it >possible that the Dutch intonation researchers, in the process of >extracting F0 targets from the intonation contours, applied >subjective criteria that somehow were affected by the common musical >tones they had been exposed to outside the laboratory? Since F0 >target extraction is not automatitized, I presume it involves >repeated listening to utterances and subsequent decision making. Is >it possible that the distributional biases arose at that stage? -------------------------------------------------------------- Alain de Cheveigne' CNRS/IRCAM, 1 place Stravinsky, 75004, Paris. phone: +33 1 44784846, fax: 44781540, email: cheveign(at)ircam.fr http://www.ircam.fr/equipes/pcm/cheveign --------------------------------------------------------------