Wanson Speech Recognition#

License#

This software is an evaluation version only. It includes a mechanism that limits the maximum number of recognitions to 50.

Overview#

The Wanson speech recognition engine runs proprietary models to identify keywords in an audio stream.

The model used in FFD is approximately 185k. The runtime and application supporting code consumes approximately 250k.

With the model in flash, the Wanson engine requires a core frequency of at least 400 MHz to keep up with real time. Additionally, the Wanson engine must be on the same tile as the flash.

To replace the Wanson engine with a different engine, refer to the FFD documentation on Replacing ASR Engine Block

Dictionary command table#

Table 17 English Language Demo#
Utterances	Type	Return code (decimal)
Hello XMOS	keyword	1
Switch on the TV	command	3
Switch off the TV	command	4
Channel up	command	5
Channel down	command	6
Volume up	command	7
Volume down	command	8
Switch on the lights	command	9
Switch off the lights	command	10
Brightness up	command	11
Brightness down	command	12
Switch on the fan	command	13
Switch off the fan	command	14
Speed up the fan	command	15
Slow down the fan	command	16
Set higher temperature	command	17
Set lower temperature	command	18

State Machine#

An optional state machine is used to condition the raw output of the Wanson speech engine.

When using the state machine, the application intent callback will only occur when a wake word and command have been detected within a time period.

Wanson speech recognizer state machine diagram

The state machine logic can be disabled by setting the compile time option appconfINTENT_RAW_OUTPUT, to 1. The wake word to command timeout is compile time configurable via appconfINTENT_RESET_DELAY_MS.

More information on these options can be found in the FFD Configuring the Firmware section.

Application Integration#

In depth information on out of the box integration can be found here: Host Integration