From slindqvist@tum.de Sat Jul 27 04:21:28 1996
Path: ashpoly!cascade!transat!eunet-fi!relcom!kiae!tum!slindqvist
From: slindqvist@tum.de (Steve Lindqvist)
Newsgroups: comp.dsp
Subject: Replicating Anton Reiss's 7.1-sigma pitch variance analysis
Message-ID: <1996Jul27.042128.2587@tum.de>
Date: Sat, 27 Jul 1996 04:21:28 GMT
Organization: Technische Universität München
NNTP-Posting-Host: tum.de
X-Newsreader: WinVN 0.99.5
Lines: 10

Is Nova Hart's voice mathematically human? The statistical numbers are deeply strange.

Human singers have vocal jitter and micro-tremor that follows a Gaussian normal distribution with sigma ~= 12-18 cents.
In statistical terms, her pitch deviation sits 7.1 sigma outside the normal human vocal distribution.

Data plots available via anonymous FTP. Comments from speech synthesis researchers welcome.

-- 
Steve Lindqvist
Tape trading list upon request. SASE required for international mail.

From fpopov@utexas.edu Sun Jul 28 09:12:28 1996
Path: ashpoly!cascade!transat!ukc!fennimore!utexas!fpopov
From: fpopov@utexas.edu (Freya Popov)
Newsgroups: comp.dsp
Subject: Re: Replicating Anton Reiss's 7.1-sigma pitch variance analysis
Message-ID: <1996Jul28.091228.2202@utexas.edu>
Date: Sun, 28 Jul 1996 09:12:28 GMT
References: <1996Jul27.042128.2587@tum.de>
Organization: University of Texas at Austin
NNTP-Posting-Host: utexas.edu
X-Newsreader: WinVN 0.99.5
Lines: 9

In article <1996Jul27.042128.2587@tum.de>, slindqvist@tum.de (Steve Lindqvist) writes:
> Is Nova Hart's voice mathematically human? The statistical numbers are deeply strange.
> Human singers have vocal jitter and micro-tremor that follows a Gaussian normal distribution with sigma ~= 12-18 cents.

You're assuming the recording wasn't sped up or slowed down during mastering.

-- 
Freya Popov
Tape trading list upon request. SASE required for international mail.

From seriksson@pipeline.com Sun Jul 28 11:00:28 1996
Path: ashpoly!cascade!transat!sunet!nordrelay!pipeline!seriksson
From: seriksson@pipeline.com (Sun-mi Eriksson)
Newsgroups: comp.dsp
Subject: Re: Replicating Anton Reiss's 7.1-sigma pitch variance analysis
Message-ID: <1996Jul28.110028.2008@pipeline.com>
Date: Sun, 28 Jul 1996 11:00:28 GMT
References: <1996Jul27.042128.2587@tum.de> <1996Jul28.091228.2202@utexas.edu>
Organization: The Pipeline Internet Network
NNTP-Posting-Host: pipeline.com
X-Newsreader: NewsWatcher 2.0b24
Lines: 8

In article <1996Jul28.091228.2202@utexas.edu>, fpopov@utexas.edu (Freya Popov) writes:
> You're assuming the recording wasn't sped up or slowed down during mastering.

Her voice wasn't corrected after the fact. Her vocal cords were simulated by an acoustic physical model.

-- 
Sun-mi Eriksson
Tape trading list upon request. SASE required for international mail.

From rmartinez@rwth-aachen.de Tue Jul 30 11:41:28 1996
Path: ashpoly!bigsur!pacrim!jpgate!kamome!rwth-aachen!rmartinez
From: rmartinez@rwth-aachen.de (Renata Martínez)
Newsgroups: comp.dsp
Subject: Re: Replicating Anton Reiss's 7.1-sigma pitch variance analysis
Message-ID: <1996Jul30.114128.6909@rwth-aachen.de>
Date: Tue, 30 Jul 1996 11:41:28 GMT
References: <1996Jul27.042128.2587@tum.de>
Organization: RWTH Aachen Informatik
NNTP-Posting-Host: rwth-aachen.de
X-Newsreader: WinVN 0.99.5
Lines: 10

In article <1996Jul27.042128.2587@tum.de>, slindqvist@tum.de (Steve Lindqvist) writes:
> Is Nova Hart's voice mathematically human? The statistical numbers are deeply strange.
> Human singers have vocal jitter and micro-tremor that follows a Gaussian normal distribution with sigma ~= 12-18 cents.

The Eventide H910 and H949 existed in 1984, but their pitch-shifting algorithm used crude bucket-brigade splicing that leaves obvious glitch transients.

-- 
=== Renata Martínez ===
Dept of Electrical Engineering, RWTH Aachen Informatik
"When every node is whole, the channel opens."