Thesis data Veronika Neumeyer: CI Articulation (31.03.2009)

F. Schiel 	09.11.2009

Questions and information about the provision of data can be directed to bas@bas.uni-muenchen.de

This corpus contains speech recordings of normal hearing
speakers and Cochlear-implant (CI) users, as used 
for analysis in the MA thesis of Veronika Neumeyer.
The data was collected with the software SpeechRecorder,
to this BPF files were generated (*.par), on which the
MAUS-Segmentation was constituted (*.TextGrid); the data 
of the selected speakers (23) was transported in '/data' and
in a Emu-database. 
The formants (fms), fundamental frequency (f0), energy (rms)
and zero-crossing rate (zcr) were calculated in Emu.
The formant values were analysed through the Emu-R gateway.


File naming

In all areas the data files were named as follows:

SSPP_WW.EXT         as   SS = speaker-ID (column SCD in SPEAEXT.TBL)
                         PP = prompt number (column 1 in PROMPTS.TBL)
			 WW = repetition 01...05
			      (spellings were different in the 
				repetitions; see PROMPTS.TBL)
			 EXT = file type (wav, f0, par, TextGrid,
				rms, fms, zcr)


Raw data

The unmodified recordings of the SpeechRecorder plus the 
adapted BPF files (orthography and canonical pronunciation)
and the TextGrid files (results of the MAUS-segmentation) are
located in the subdirectories in '/rawdata'.
The naming of the subdirectories corresponds to the column
'FNR' in the table '/table/SPEAEXT.TBL'.
Each directory corresponds to one speaker.
iSpeech signals are stored in RIFF WAVE format, 1 channel, 22050Hz
sampling rate, 16bit PCM (little endian). Recording technique::
MAUDIO Mobile Pre or Yamaha Digital Mixer connected to PC (WinXP), 
recording software SpeechRecorder, microphone headset Sennheiser USB36.

Emu data

The directory '/data' contains the data of the Emu-Database
in the subdirectories f0, fms, rms and zcr.
The Emu template is provided in '/data/table/ci-sprecher.tpl'.
The template is already conformed for this installation; for
getting access via Emu only the path '/bmnt/BAS/IPS-PROJECTS/table
has to be entered in the configuration editor of Emu.

The course of the formants were manually checked at the vowel
in the first syllable of the target word and if
necessary corrected.
Default-adjustments of Emu were used for the calculation of the
derived signals.

The automatically created Emu annotations (in '/annot') were
generated out of the corrected TextGrids. The particular types of
Emu annotation are located in the subdirectories hlb, ORT, KAN, MAU.
The subdirectory textgrid contains the corrected TextGrid files of
the MAUS-segmentation, which then were loaded in Emu.
Before the loading the following corrections were done with praat:
- word boundaries of the target word
- at boundaries of the vowel in the first syllable of the target word

Tables

Meta data of the recordings and the speakers are located in '/table'.
Speakertable SPEAEXT.TBL contains the columns:
SCD = speaker number
FNR = folder number
SEX = M|F
AGE
ACC = accent = German federal land where the primary school is located
   BW : Bademwuertemberg
   BY : Bayern
   NO : ???
   NI : Niedersachsen
   NW : Nordrheinwestphalen
   RP : Rheinlandpfalz
   SN : Sachsen
NAL = country of origin and native language, ISO3166-1 code
SGR = speaker group (CI or CG (control group))
H-I = (kind of) hearing impairment (deaf or hearing impaired?)
H-A = hearing aid (yes or no?)
SIL = (use of) sign language (yes or no?)
CIS = CI-supply (unilateral vs. bilateral)
AAI = age at implantation
LOU = length of use
The prompt-table PROMPTS.TBL contains 2 columns
PP PROMPTSTEXT
The prompts 01 - 20 contain simple sentences, which were repeated 5
times (WW = 01...05). The prompts 21 and 22 contain spellings
(coded as capital letters), which were different in all repetitions.
According to this the table PROMPTS.TBL contains the entry PP_WW
in the first column.

Documentation

This file and the MA thesis as PDF are located in '/doc'.

Any other business

Speaker 82 (recording 0020) is incomplete.

For analysis in the thesis following speakers were used:
11,12,13,16,17,18,19,22,23,24,25,26,27,35,57,68,72,75,80,82,91,95,R5
The remaining speakers were not annotated (only in '/rawdata').


Contact:
Name: Florian Schiel
Address: Institut für Phonetik, Schellingstraße 3/II, 80799 München
Email: bas@bas.uni-muenchen.de
Organisation: Bavarian Archive for Speech Signals, Ludwig-Maximilians-Universität München
Telephone: +498921802758
Website: http://hdl.handle.net/11858/00-1779-0000-000C-DAAF-B