![NXP Semiconductors SLN-LOCAL2-IOT User Manual Download Page 40](http://html.mh-extra.com/html/nxp-semiconductors/sln-local2-iot/sln-local2-iot_user-manual_1721901040.webp)
Chapter 7
Far-field local voice control framework
Figure 46. High-level overview of far-field local voice control framework
This section describes the software framework that supports the far-field local voice control. As shown in
, two (optionally
three) microphones collect the acoustic signal, followed by the DSP, AFE, and ASR blocks.
The SLN-LOCAL2-IOT kit is acoustically qualified for far-field voice applications with three PDM microphones and has been
internally tested with two-microphone configurations with a range of mainstream products also using the two-microphone
configuration. When making modifications, ensure to re-test the application against standard acoustic test guidelines. The
SLN-LOCAL2-IOT kit is based on the acoustic architecture of the SLN-ALEXA-IOT kit. It was tested based on the Amazon Voice
Service self-test guidelines, which are available at
.
NXP has pre-tuned and qualified the DSP and AFE libraries with the SLN-LOCAL2-IOT hardware platform. By default,
modifications on the DSP and AFE are not needed. However, to create customized hardware or proof-of-concepts, see
and ensure that the modification is suitable for your product.
contains the speech recognition engine and the application software. NXP has implemented the
following three types of baseline demos:
• LED voice control demo
— English
— Two-stage (wake word and command) ASR
• Smart Home (IoT) or elevator or audio device or washing machine voice control demo
— Selectable combinations of English, Chinese, German, and French
— Two-stage ASR
• Oven voice control demo
— English
— Multiturn (4-way) dialog-style ASR
The ASR implemented with the selected languages can be easily replaced with other languages. NXP provides an application
note for customization of the local voice demos. Contact NXP (
) for information about the process of
phoneme-based speech recognition engine generation and custom wake words and commands.
describes the baseline ASR demos that you can reuse for your product.
7.1 Automatic speech recognition
The flagship feature of SLN-LOCAL2-IOT is the bundled voice control engine, also called ASR. NXP offers a lightweight engine
designed specifically for MCUs. It supports various use cases with flexible inference engine instances.
NXP Semiconductors
SLN-LOCAL2-IOT Developer’s Guide, Rev. 0, 19 April 2021
User's Guide
40 / 87