Hello,
I bought the Respeaker 4-Mic array to use it in my project. I want to say orders and the respeaker convert the order to text so the python code that i have would do the order i asked for. The problem is that there isn’t much info about how to use it like that in the internet. How to make the respeaker recognize my key words?
please help.
Hi,
There are some readymade things for that you can use with the Respeaker like Amazon Alexa, Google voice assistant, you can build on top of these frameworks otherwise you need to implement your own NLP system from scratch of available python libraries .
Hi,
Thank you for your reply. Could you direct me to links that teach how to build a NLP system from scratch of available python libraries? i am not an expert on python.
Sure,
Alexa: https://github.com/respeaker/avs
Google Assistant: <LINK_TEXT text=“https://aiyprojects.withgoogle.com/voic … -assistant”>https://aiyprojects.withgoogle.com/voice/#google-assistant</LINK_TEXT>
These are ready made systems.
not sure you’re getting much to go on from their links. So I can try to tell you what I’ve figured out so far which is kinda a bewildering lack of coherent information. There are 2 different parts to a voice system, there’s a independent part for key word detection (aka wake word detection) and then a different part for making sense of the voice stream. In most of the Seeed tutorials they’re using snowboy as the hotword detection system. Snowboy is fairly dated and hasn’t been updated in quite awhile and it only works with python 2.7 so for numerous reasons it may not be the best basis for anything you want to do commercially. There is a system called Porcupine which I haven’t messed with a lot and then one from Mycroft called Precise. Precise is the one I’ve been trying to use and seems like the best platform from what I’ve read for development of your own product - its both opensource but also is getting the necessary love and attention because it is integrated into multiple commercial systems. Then for the streaming service you have a couple main offerings - those from google and amazon but both of these require internet connectivity. Amazon’s system seems a little easier to use and a bit more built out (to me) but I think google’s approach to allowing local fulfillment has some advantages if you’re trying to interact with a local product… obviously both of these companies are trying to win the “land grab” phase of this market and will try to capitalize on them at some point in the future. For my own needs i’m looking particularly at DeepSpeech as i’d prefer not to have the alexa or google home connectivity requirement. So, given the development that’s already gone into amazon and google’s approaches i’m utilizing those as my “starter” system with the intent to eventually replace them with a hoped for deepspeech based intent routing system at some point in the future. Anyway, hopefully that gives you a little more to go on and or a few more things to look at.
1 Like