Page 1 of 1

Adding custom voice recognition commands on Esp32-Lyra T .

Posted: Fri May 17, 2019 1:37 pm
by vikneshwar
# 1. Can we recognize a wide range of English words? (We only found some examples of Chinese commands)
# 2. Whether ESP32 is compatible with automotive applications.
# 3. Support offline speech recognition using custom commands (name, command)
# 4. How to add custom voice commands.

Re: Adding custom voice recognition commands on Esp32-Lyra T .

Posted: Sat May 18, 2019 8:59 am
by ESP_Sprite
The speech recognition engine we have unfortunately does not allow free-form speech to be recognized. If you want to use this for a commercial application, I'd suggest you contact sales@espressif.com and we'll see what we can do for you.

Re: Adding custom voice recognition commands on Esp32-Lyra T .

Posted: Sat Sep 21, 2019 12:03 pm
by heinzv
I have also purchased 4 ESP32 audio boards (LyraT and ESP31-A1S Audio Kits from AI) with the idea to have local speech recognition for at least 20 commands.
Unfortunatly we get only a hadnfull useless Chineese Words. So it is totally useless outside China.
It would only makes sense if a handfull English (in my case even German) words could be trained. I have no use for Alexa online recognition.

This is NOT for commercial usage.

Re: Adding custom voice recognition commands on Esp32-Lyra T .

Posted: Fri Feb 07, 2020 10:36 am
by sergiomarina
Hello,
i'm interested in creating a business application that uses speech recognition.
So i'm wondering if ESP32-Lyra can do it. The drafted application is as follows:
1) user tells wake up word (whatever made available by API) to Lyra
2) Lyra answers "ok" (or whatever user-recorded or pre-recorded) if recognition was successful OR Lyra answers "ko" (or whatever) if it dis not recognize (if so the user needs to start back to step 1)
3) user speech is made by 6 words - three single digits [0-9] (in local language)+ the word "plus" o "minus" (the translation in local language) + two single digits [0-9] (in local language)
4) Lyra records the 6 events
5) Lyra answers "ok" (or whatever) if recognition was successful OR Lyra answers "ko" (or whatever) if it did not recognize (if so the user needs to start back to step 1)
6) Lyra stores the message in RAM and the application processes related data accordingly to the Customer specifications

Is it possible to implement this procedure using off the shelf APIs? (I have recently installed ESP-IDF version 4.1)

If the answer is "yes", I'd like to if other ESP boards can achieve that result.
If the answer is "no", I'd like to weather Espressif offers a viable solution for that.

Thank you and best regards, Sergio