search

MrBuddyCasino / ESP32_Alexa

An Alexa Smart Speaker project for the ESP32.
Mozilla Public License 2.0
264 stars 72 forks source link

Wake word support #21

Open jonsmirl opened 6 years ago

jonsmirl commented 6 years ago

Espressif has a wake word engine that runs on the ESP32 now. It works with Alexa.

But... it needs that module with 4MB of PS RAM.

hchaudhary1 commented 6 years ago

open source?

jonsmirl commented 6 years ago

I don't know... this is from Espressif

Peter Wang | Peter Wang | 12:33 PM (10 hours ago) |   |  

We have got the IA610 EV board, will make a prototype board for the solution. If SPH0645LM4H-B be used for near field application , I think it is ok.

The wake word can be done on ESP32 or External DSP. The AEC function should be done on external DSP , if use a Microphone array. If it is done on ESP32 , then main CPU of ESP32 should keep working to detect and process the wake word. If it is done on external DSP, then ESP32 can enter sleep mode to saving the power. That would be useful for battery power device.

jonsmirl commented 6 years ago

It is part of this kit. But that $10 DSP is not required.

https://www.cnx-software.com/2018/01/16/espressif-esp32-lyratd-ms1-hdk-is-designed-for-smart-speakers-wireless-audio-and-other-smart-home-appliances/

jonsmirl commented 6 years ago

It need the module with the 4MB PSRAM - ESP32-WROVER https://www.espressif.com/sites/default/files/documentation/esp32-wrover_datasheet_en.pdf

hchaudhary1 commented 6 years ago

no code released yet... keeping an eye out for it. would be helpful if they don't need to use the $10 microsemi part

jonsmirl commented 6 years ago

Peter said... If SPH0645LM4H-B be used for near field application , I think it is ok.

That is $0.50 part.

The Lryatd kit was released in January. Ping the CNX guy, he has the kit Jean-Luc Aufranc cnxsoft@cnx-software.com

shirish47 commented 6 years ago

I am also looking for this functionality wakeup word detection. One way I think it can be done is by HM2007 chip(just detecting word and waking ESP32 to send other part of the voice).

yoelrc88 commented 6 years ago

Has someone tried to use the Cloud-Based Wake Word Verification to achieve this? I wouldn't mind sending an audio stream to Amazon servers ... at least for testing :) Thanks

MrBuddyCasino commented 6 years ago

yoel: thats just verification to check if you were wrong or not, you still need to do wakeword detection

danxster commented 6 years ago

@shirish47 - there's also: EasyVR Shield 3.0 ELECHOUSE Voice Recognition Module V3 Tigal SmartVR Voice Recognition Board Grove – Speech Recognizer

jonsmirl commented 6 years ago

Espressif LyraTD SDK code is here: https://github.com/espressif/esp-adf

Not sure if the wake word support is in the SDK yet.

Looks like a beta is here: https://github.com/espressif/esp-avs-sdk/releases

https://github.com/espressif/esp-avs-sdk/blob/master/README.md