Cirrus Logic CS48LV13 - Manual (Page 11)

11 DS1057F1
3.13 Speech DSP Chain
3.13 Speech DSP Chain
The CS48LV13 includes Cirrus Logic ASR Enhance preprocessor that removes noise impacting ASR engine
performance. SoundClear ASR Enhance algorithms use specialized spatial NR techniques to improve command success
rate by accomplishing the following:
• Reducing near-end noise that masks speech pauses and otherwise interferes with accuracy
• Suppressing cross-talk from competing talkers in the vicinity
• Preserving voice spectral content
• Preventing phoneme degradation
• Using InstantAdapt™ noise adaption to avoid missed or truncated commands
The ASR Enhance preprocessor can be used to improve local ASR performance including the optional Sensory
TrulyHandsfree voice control as well as cloud based engines.
The CS48LV13 supports an optional speech chain, which can include the following functions:
• Cirrus Logic voice activity detector (VAD). This specialized VAD enables always-on ultralow-power voice trigger.
The VAD constantly monitors ambient sound while the CS48LV13 is in a very low power state. If it detects voice,
the VAD can trigger the supported Sensory TrulyHandsfree voice control engine running on the CS48LV13.
• Sensory TrulyHandsfree voice control. When the VAD has detected a voice, it can trigger TrulyHandsfree to detect
a predefined wake-up command. If it detects the proper command, it enters Voice-Command Mode to interpret any
command within its vocabulary. If the proper wake-up command is not detected, the device can return to its
always-on VAD mode.
3.14 On-chip DSP Peripherals
3.14.1 I 2 S Digital Audio Ports
3.14.1.1 I 2 S Digital Audio Input Ports (DAI)
Two DAI ports support PCM format with word lengths up to 32 bits and sample rates as high as 192 kHz. DAI_1 is typically
connected to a host processor, applications processor, or mobile SoC to receive Rx voice and audio data. DAI_2 is
typically connected to a codec or A/D streaming Tx voice or audio data captured from microphones. Both ports operate in
I
2 S slave clock mode using SCLK and LRCLK for bit-clock and word select. For voice (call) modes, both DAI ports use a
16-bit word length, a 8-KHz sample rate for narrowband calls, and 16-KHz sample rate for HD voice/wide-band calls. Rx
(downlink) requires one channel and Tx (requires one or two channels), depending on microphone/input mode. For
playback mode, other audio formats are supported, based on the content type and postprocessing algorithms applied.
3.14.1.2 I 2 S Digital Audio Output Port (DAO)
Two DAO ports support PCM format with word lengths up to 32 bits and sample rates as high as 192 kHz.
DAO_1 is typically connected to a host processor, applications processor or mobile SoC to transmit Tx voice and audio
data. DAO_2 is typically connected to a codec, D/A, or digital amp streaming Rx voice or audio data out to transducers
such as mobile phone receiver, speakers, or headphone. Both ports operate in I
2 S slave clock mode using SCLK and
LRCLK for bit clock and word select. For voice (call) modes both DAO ports will use a 16-bit word length, 8-KHz sample
rate for narrowband calls, and 16 KHz for HD voice/wide-band calls. For playback mode, other audio formats are
supported based on the content type and postprocessing algorithms being applied.
3.14.2 Serial Control Port (I 2 C or SPI)
The on-chip serial control port is capable of operating as slave in either I 2 C or SPI modes. Slave operation is chosen by
a mode select pin when the CS48LV12/13 comes out of reset. The serial clock pin can support frequencies as high as
25 MHz in SPI mode.
The slave SPI clock speed must always be (DSP Core Frequency/2).