
Preliminary
DSP for Digital Answering phone with FLASH interface
KS16121
FUNCTIONAL DESCRIPTION
HOST INTERFACE
The KS16121 acts as a co-processor to a host. It communicates with the host via an 8-bit parallel interface.
A simple protocol of 16-bit instruction issued by the host and a 16-bit status word returned from the KS16121
is employed. Through this command / status protocol, the host can handle the KS16121 to perform multiple
functions simultaneously in various modes.
MESSAGE RECORDING
The KS16121 uses 4M( 512K x 8 ) or 16M (2M x 8 ) bit SAMSUNG NAND flash memories to store the
compressed speech data. A recording operation generates a compressed representation of speech segment, or
compressed message, and saves it in the flash memories. Subsequent recording operations store the compressed
messages sequentially in the memory space until no more space is available. A maximum of 128 messages
can be stored.
The KS16121 employs a software algorithm running on an on-chip DSP core to compress the incoming speech
samples at 8 KHz time intervals from a
μ
-law PCM codec. The algorithm processes the speech on a 20 msec
time frame and produces the compressed data at 5.15 Kbps. When used with the silence compaction option,
the algorithm can store approximately 15 minutes of compressed speech in a 4M flash memories.
The silence before a speech signal or between speech segments can be detected and used by the KS16121
to save the memory space needed to store the message. The host may choose the following recording options:
Voice Activation
--
The initial silence is ignored and the recording starts only when the KS16121 detects a
speech signal. Without this option, the recording starts immediately after a record instruc-
tion.
Silence Compaction
--
The silence gaps between speech segments are measured and replaced with simul-
ated noise during playback, to save the memory space.
A recording operation normally ends with a host command set . When necessary, the length of the
message may be reduced by removing the tail end of the message, which may contain no speech signal. The
instruction from the host specifies the number of speech frames to be removed.
MESSAGE PLAYBACK
A playback operation retrieves a compressed message from flash memories. Each message is identified with a
message number, given by the host along with a playback command. A portion of memory space is allocated
to maintain the information related to individual messages, such as memory address and message length.
The KS16121 reads the compressed message from flash memories and processes it with the decoding software
algorithm to recover the original speech samples. The samples are fed to an external
μ
-law PCM codec for
digital to analog conversion.