NSAM266SA Datasheet, PDF (25/52 Pages) National Semiconductor (TI) – NSAM266SA CompactSPEECHTM Digital Speech Processor with Serial Flash Interface

2 0 Software (Continued)
A message is the basic unit on which most of the Compact-
SPEECH commands operate A CompactSPEECH mes-
sage stored on a flash device can be regarded as a com-
puter file stored on a mass-storage device
A message is created with either the R or the WRAM (Write
Memory) command
When a message is created it is assigned a time-and-day
stamp and a message tag which can be read by the micro-
controller
The R command takes voice samples from the codec com-
presses them and stores them in the message memory
When a message is created with the WRAM command the
data to be recorded is provided by the microcontroller and
not via the codec The data is transferred directly to the
message memory It is not compressed by the Compact-
SPEECH voice compression algorithm
The WRAM command together with the RRAM (Read
Memory) command which enables the microcontroller to
read data from the CompactSPEECH can be used to store
data other than compressed voice in the message memory
e g a telephone directory
A message can be played back (P command) and deleted
(DM command) Redundant data (e g trailing tones or si-
lence) can be removed from the message tail with the CMT
(Cut Message Tail) command
The PA (Pause) and RES (Resume) commands respective-
ly temporarily suspend the P and R commands and then
allow them to resume execution from where they were sus-
pended
CURRENT MESSAGE
Most message handling commands e g P DM RRAM op-
erate on the current message The GTM (Get Tagged Mes-
sage) command selects the current message
Deleting the current message does not cause a different
message to become current The current message is unde-
fined If however you issue the GTM command to skip to
the next message the first message that is newer than the
just deleted message is selected as the current message
2 6 1 Message Tag
Each message has a 2-byte message tag which you can
use to categorize messages and implement such features
as OutGoing Messages mailboxes and different handling
of old and new messages
The most significant bit (bit 15) of the message tag is used
to indicate the speech compression rate The microcontrol-
ler should program it before recording (‘‘1’’ for 4 8 kbit s
‘‘0’’ for 6 6 kbit s) The CompactSPEECH reads the bit be-
fore message playback to select the appropriate decom-
pression algorithm
The GMT (Get Message Tag) and SMT commands may be
used to handle message tags
Note Message tag bits can only be cleared Message tag bits are set only
when a message is first created
This limitation is inherent in flash memories which only allow bits to
be changed from 1 to 0 (changing bits from 0 to 1 requires a special
erasure procedure see Section 1 3 5) However the main reason for
updating an existing tag is to mark a message as old and this can be
done by using one of the bits as a new old indicator setting it to 1
when a message is first created and clearing it when necessary
2 7 SPEECH COMPRESSION
The CompactSPEECH implements two speech compres-
sion algorithms One algorithm with 5 2 kbit s compression
rate enables up to 14 – 16 minutes of recording on a 4-Mbit
device while the other uses a 7 3 kbit s compression rate
to support 10 – 12 minutes of recording Both compression
rates assume 10% silence
Before recording each message the microcontroller can se-
lect one of the two algorithms by programming bit 15 of the
message tag During message playback the Compact-
SPEECH reads this bit and selects the appropriate speech
decompression algorithm
IVS vocabularies can be prepared in either of the two com-
pression formats using the IVS tool All the messages in a
single vocabulary must be recorded using the same algo-
rithm (See the IVS User’s Manual for further details) During
speech synthesis the CompactSPEECH automatically se-
lects the appropriate speech decompression algorithm
2 8 TONE AND NO-ENERGY DETECTION
The CompactSPEECH detects DTMF busy and dial tones
and no-energy (VOX) This enables remote control opera-
tions and call progress Detection is active throughout the
operation of the CompactSPEECH Detection can be con-
figured using the SDET (Set Detectors Mask) command
which controls the reporting of the occurrence of tones and
the RDET (Reset Detectors) command which resets the de-
tectors
DTMF
DTMF detection may be reported at the starting point end-
ing point or both The report is made through the status
word (for further details see GSW command in Section
2 14)
The DTMF detector performance as measured on the line
input using an NSV-AM265-DAA board is summarized be-
low (see Table 2-1)
ECHO CANCELLATION
Echo cancellation is a technique used to improve the per-
formance of DTMF tone detection during speech synthesis
tone generation and OGM playback For echo cancellation
to work properly AGC must not be active in parallel Thus
to take advantage of echo cancellation the microcontroller
must control the AGC i e disable the AGC during PLAY
SYNTHESIS and TONE GENERATE states and enable it
again afterwards If AGC cannot be disabled do not use
echo cancellation The microcontroller should use the CFG
command to activate deactivate echo cancellation (For fur-
ther details see Section 2 14 )
Echo cancellation applies only to DTMF tones Busy and
dial-tone detection is not affected by this technique This
implies that the performance of the busy and dial-tone de-
tector during message playback depends on the message
being played
25
http www national com