TrueSpeech® Co-Processor
Notes:
PRELIMINARY/CONFIDENTIAL
Version: 1.18
1. The CT8022 supports the FlexiSpeech, G.729A+B and G.722 speech coders as external downloadable
expansion software modules. These are not built-in to the CT8022 internal program ROM. Use of
downloadable expansion software modules requires that the CT8022 be used in conjunction with the
appropriate application circuit described in Section 2.2 of this data sheet (with 32K x 16 external
SRAMs). All other speech compression functions (G.723.1, TrueSpeech 8.5, 4.8, and 4.1, G.711 and the 8-
bit & 16-bit uncompressed modes) are built-in to the CT8022.
3.3
TrueSpeech
TrueSpeech is a speech compression technology that reduces or compresses the amount of data used to encode a
speech waveform. TrueSpeech compression reduces speech sampled (digitized) at 8,000 samples per second, in 16-
bit samples (128,000 bits/sec) to 8.5, 6.3, 5.3, 4.8 or 4.1 Kbits/sec with minimal degradation in speech quality.
When used with an external G.711 A-law/µ-law CODEC, the CT8022 provides intermediate conversion from G.711
A-law/µ-law to 16-bit linear format prior to the operation of the TrueSpeech function
In the CT8022, uncompressed speech is obtained from an external CODEC operating at 8,000 samples per second.
If an external G.711 A-law or µ-law CODEC is used, the speech is sampled by the external CODEC with 14-bits of
effective resolution. The G.711 CODEC converts the 14-bits (linear) samples to 8-bit A-law/µ-law for transmission
to the CT8022 (at 64,000 bits/sec). The CT8022 receives the speech samples from the CODEC and converts them
back to 14-bit linear form. The 14-bit data is converted to 16 bits by appending two zero bits in the least significant
bit positions. The 16-bit data is collected into blocks of 240 samples (480 byte) every 30ms and TrueSpeech
compressed for transfer to the Host. If an external linear 16-bit CODEC is used, the speech data is sampled with 16-
bit resolution and passed directly from the CODEC to the CT8022 at 128,000 bits/sec.
The CT8022 supports simultaneous real-time compression and decompression of TrueSpeech data.
3.4
TrueSpeech 8.5
In TrueSpeech 8.5 mode, speech data is compressed/decompressed into 16-word (32 byte) blocks, which are
transferred to or from the Host every 30ms. This corresponds to a data rate of 8,533.33 bits/sec. The CT8022
TrueSpeech 8.5 implementation is compatible with the TrueSpeech 8.5 format used in the CT8005 TrueSpeech
Messaging Co-Processor, CT8015 TrueSpeech DSVD Co-Processor and Microsoft Windows Sound System 2.0 and
Windows 95 software TrueSpeech CODECs.
3.5
TrueSpeech 6.3 (G.723.1)
The CT8022 implements the final (version 5.1) ITU-T formal release of the G.723.1 speech coder standard,
including G.723.1 Annex A VAD/CNG.
In TrueSpeech 6.3 (G.723.1), data is compressed into 12-word blocks. This provides 192 bits per block. However,
only 189 bits are used for actual speech data, giving a raw data rate of 6,300 bits/sec. When used in compliance with
G.723.1 two of the spare bits are assigned for control information (the 2 least significant bits of the first word of each
compressed speech block). The CT8022 G.723.1 implementation includes the optional G.723.1 Annex A
VAD/CNG silence compression feature (Voice Activity Detection/Comfort Noise Generation). When this feature is
enabled, the CT8022 can generate and decode the 4-byte and 1-byte Annex A silence frames. The CT8022 supports
dynamic switching between the G.723.1 6.3 and 5.3 rates for both the encoder and decoder. In addition, the G.723.1
6.3 and 5.3 decoders support frame erasure (for corrupt or missing frames) triggered using specially marked data
frames.
20
DSP GROUP, INC., 3120 SCOTT BOULEVARD CT8022A11AQC FW Revision 0118
SANTA CLARA, CA 95054 PH: 408 986 – 4300 FAX: 408 986 – 4490
All specifications are subject to change without prior notice.
ETC [ ETC ]
