Speech Data Processing Documents 990

[+FarmerPotato]

Speech ROM Data Preparation Software System

 

Gene Helms

Texas Instruments
April 1979

 

This is a user's reference guide for the 990 tools which digitize speech, create LPC, and output ROM data.

 

 

DIG990 captures analog to digital. 
DSEDIT segments the recording into one file per phrase (DS) and creates a phrase library. 
LPCANA produces coded speech parameters (CS)
LPCOPT optimizes for repeated frames (FR)
CSEDIT edits CS or FR data, or outputs synthesized speech
CODPAK quantizes for a coding/decoding table, and creates ROM data.

 

 

Speech ROM Data Preparation Software Apr1979.pdf

[+FarmerPotato]

Memo

 

Instructions on how to use the new DSDD floppies, in the DX10 operating system for the 990.

 

Jay Reimer

Texas Instruments

April 13, 1982

 

 

Memo_Instructions_for_DSDD_disks_Jay_Reimer_Apr1982.pdf

[+FarmerPotato]

Directory listing of 8" backup floppy disk.

Home directory of KMG, containing source to the Text-To-Speech software. (KMG = Kathleen M. Goudie)

 

I have a Shugart SA-850 on the way, so I will try to read these floppy disks. 

 

Directory_Listing_KMG_1983_12_16.pdf

[jbdigriz]

I could swear there was an article describing this system in some issue of 99'er, but I can't seem to find it right now.

 

I would consult with bitsavers or one of the other people active in pulling data off old 8" floppies, if you haven't done so already. There are various treatments and procedures that will at least get you one good pass in cases where oxide is shedding, there's mold/mildew, etc. Hopefully these disks were stored cool and dry. Regrettably I can't say the same for most of mine.

 

Again, good luck with it. Very interested to see what you find.

[+FarmerPotato]

2 hours ago, jbdigriz said:

I could swear there was an article describing this system in some issue of 99'er, but I can't seem to find it right now.

 

I would consult with bitsavers or one of the other people active in pulling data off old 8" floppies, if you haven't done so already. There are various treatments and procedures that will at least get you one good pass in cases where oxide is shedding, there's mold/mildew, etc. Hopefully these disks were stored cool and dry. Regrettably I can't say the same for most of mine.

 

Again, good luck with it. Very interested to see what you find.

I believe the disks have been in this binder since 1984, in a cool, dry place.  Unmoved for 36 years.

 

It occurred to me today that "KMG" are the initials of Karl M. Guttag.  It seems unlikely that he was in Speech Products. But perhaps I should try to return the disks.

 

 

[+FarmerPotato]

Voice Evaluation

 

George Doddington, Richard Wiggins
Texas Instruments
22 January 1979

 

This lays out the factors for evaluating human speakers. 

 

 

 

Voice_Evaluation_Jan1979.pdf

[+FarmerPotato]

Available Design Options in Speech Synthesis Systems Based on the TMCO280

 

Richard Wiggins
Texas Instruments
20 February 1979

 

This memo lays out considerations for a speech chip to come after the TMC0280 (Speak and Spell.)

 

Design_Options_in_Speech_Synthesis_TMCO280_Feb1979.pdf

[jbdigriz]

32 minutes ago, FarmerPotato said:

I believe the disks have been in this binder since 1984, in a cool, dry place.  Unmoved for 36 years.

 

It occurred to me today that "KMG" are the initials of Karl M. Guttag.  It seems unlikely that he was in Speech Products. But perhaps I should try to return the disks.

 

 

Yeah, seems unlikely, but you never know. I doubt he'd mind your archiving the disks if so, but wouldn't hurt to reach out. Might could tell you more about them, if it's him. Might want to wait till you've had a chance to look at them, though. Probably not the only "kmg" to work for TI in that era.

 

 

 

 

[jbdigriz]

Or it could be a project acronym or something. ...nah, doesn't really seem to fit. It does look like the base part of a DX10 username, though.

Edited September 11, 2022 by jbdigriz

[+FarmerPotato]

Proposed general purpose 5100/5200 coding/decoding tables
 

Mike McMahan
20 Septemer 1980
Texas Instruments

 

Data to support the choice of coding table for the "Home Computer" speech chip. 
It contrasts the proposed coding table with TI's prior chips and non-TI efforts. 

 

Of interest here:

 

The 0280's excitation or "chirp table"  had DC bias removed (for Speak and Spell), while the 0285 was DC biased.
The 2801 shared the 0280's pitch table of 31 steps. 
The 2802 increased the range of pitch. 
The 0285 doubled the number of pitch steps to 63.

 

Elsewhere:

 

TMC0285 - Schematics dated 2 November 1978.

The "Home Computer" speech chip, also known as 9928 or "mod to 0280", had been scheduled for first prototype in March 1979 (in a plan dated September 1978.)
 

 

 

Proposed_5100_5200_coding_decoding_tables.pdf

[jbdigriz]

From @kl99's xlnt interview with Herman Schuurman:

 

Q) Can you describe in detail your involvement with the TMS5200 speech synthesizer?

 

The original design for the TMS5100 was done by Larry Brantingham, Paul Breedlove, Richard Wiggins, and Gene Frantz. Gene was heading up the speech group (home of the Speak & Spell) when I joined TI in Lubbock, and Larry moved to TI France (Nice) soon after.  I eventually wound up in a group with Richard Wiggins when I joined the Corporate Engineering Center in Dallas.

The second generation of the TMS5100, the TMS5200, was used to power the speech peripheral. My involvement with the speech synthesizer was to create the text translation and allophone stringing software in coordination with Kathy Goudie (who worked for Gene), who was responsible for creating the allophone (sound) set and the translation rules.  The linked article by Sharon Crook is basically a rehash of the internal documentation on the text to speech software.

 

So, Kathy Goudie is a possible, maybe.

 

Full interview at https://forums.atariage.com/topic/295223-historical-interviews-with-ti-employees/

 

Edited September 11, 2022 by jbdigriz

[+FarmerPotato]

1 hour ago, jbdigriz said:

Herman Schuurman:

 

My involvement with the speech synthesizer was to create the text translation and allophone stringing software in coordination with Kathy Goudie (who worked for Gene), who was responsible for creating the allophone (sound) set and the translation rules.  The linked article by Sharon Crook is basically a rehash of the internal documentation on the text to speech software.

 

 

I am so dense.  Of course.  Kathleen M. Goudie (KMG) was the linguist in the speech lab.  She was co-author of 2 IEEE articles with Gene Frantz, which I read recently. Thank you for the clarification. 

 

"Text-To-Speech Using LPC Allophone Stringing".
Kun-Shan Lin, K. M. Goudie, G. A. Frantz,  G. L. Brantingham.
IEEE Transactions on Consumer Electronics, May 1981.

"Application of Synthesis-By-Rule System to Singing". Gene Frantz, Kun-Shan Lin, Kathleen M. Goudie.
IEEE Transactions on Consumer Electronics, August 1982.

 

(Note: the article says that a pitch table covers 3 tempered octaves (36 code points) within 2% error. Pitch 1 of the table is 500 Hz and pitch 56 is 65 Hz, at the 8 kHz sample rate.)
 

 

[jbdigriz]

Cool 👍

[+FarmerPotato]

LPC Speech Processing System Manual

Albert J. Olson
June 16, 1983
Texas Instruments

User's manual.  Contains some history of speech lab systems on TI 980 and TI 990 computers.  Commands for playing speech from the phrase library, and editing it frame by frame.  Glimpses into the speech data files, seen with SDLIB.

 

This document is special because it was written by my Dad.  Presented to Texas Tech University for students to use, on a 990 installed by TI. 

 

 

LPC_Speech_Processing_System_Manual_Jun1983.pdf

[+FarmerPotato]

Checking the library at University of Texas, Dallas, 2013. 

Lots of old books about speech processing.  

Went there for an IEEE lecture. 

 

Hess, Wolfgang.  Pitch Determination of Speech Signals.  Springer, 1983.