ESP32-Korvo AI Development Board Launched

Shanghai, China
Apr 23, 2020

ESP32-Korvo is Espressif’s new AI development board. Equipped with a multi-microphone array, it can achieve high performance, spoken-command recognition and far-field voice wake-up.

ESP32-Korvo is an AI development board for embedded IoT devices. It is based on ESP32 and ESP-Skainet, Espressif’s speech recognition SDK. ESP32-Korvo features a multi-microphone array which is suitable for far-field speech recognition applications with low-power consumption. ESP32-Korvo is composed of two boards connected with an FPC cable. The overall board integrates multiple peripherals, earphones, speakers, an SD card slot, LED lights, and function buttons, all of which can serve a variety of AI-powered voice applications.

In recent years, with the rise of Artificial Intelligence and the Internet of Things, voice commands have become a fairly common way for users to interact with smart devices. ESP32-Korvo is equipped with a single ESP32 chip and supports low-power, far-field speech recognition applications.

ESP32-Korvo functions with front-end acoustic algorithms, such as microphone-array speech enhancement (also known as MASE) and acoustic echo cancelation (AEC), which effectively reduces the interference of any noise in far-field speech commands, while improving the performance of voice wake-up and speech recognition.

Comparison before (top graph) and after (lower graph) the noise reduction effect of ESP32-Korvo on a sample taken from a crowded environment

The motherboard of ESP32-Korvo contains the ESP32-WROVER-B module (16 MB flash, 8 MB PSRAM), a power port, a micro-SD card slot, as well as earphone and speaker connectors. The daughterboard contains a microphone array, function buttons and LEDs.


ESP32-Korvo diagram

To achieve optimal application development, users should combine ESP32-Korvo with Espressif’s Smart Voice Assistant, ESP-Skainet. Its voice wake-up engine, WakeNet, enables users to customize wake-up words. Also, ESP-Skainet’s command recognition network, MultiNet, enables users to easily set a variety of command words, add their own voice commands, and realize local (namely, not cloud-based) voice control over smart devices.

In the near future, ESP-Korvo will also include functions such as sound-source localization (direction-of-arrival detection) and reverberation cancelation (de-reverberation). So, stay tuned for all our subsequent updates on ESP-Skainet.

Share this article
Reuse this content