Original link: https://atjason.com/daily/2023-07-23.html
I have always been interested in smart speakers, and I tossed again on the weekend.
After a brief survey, the existing smart speakers are very closed, with very limited opening capabilities. If you want to re-develop the existing speakers, you can’t achieve the desired effect.
If you start from 0, it will take too much time and effort.
The compromise method is to find a formal technical solution, and modify and adapt it on this basis to meet your own needs.
After a brief research, I chose the open source project https://wukong.hahack.com/ (I have a brief understanding of the author, and I may be about the same age as myself, ashamed). Simply install the operating environment on the macOS system, and it basically works: support custom wake-up words, recognize speech to text, request from OpenAI and get answers, convert text to speech and play it.
The next step is to move this system to a small development board and package it into a speaker. I have a dust-eating Radxa Rock in my hand, I took it out to blow off the dust, turned it on, and it started. Tried to install, ran into all kinds of problems until it was almost impossible to solve. The core is that the hardware is too old, almost 10 years ago, and the software system is too old, it is not easy to upgrade to the latest; and the Wukong software requires a relatively new operating environment. The two don’t match, so it won’t work.
They all want to spend more money and buy a development board that can run the latest version of Ubuntu. This leaves two options:
- raspberry pie. The advantage is that the software ecosystem is complete; the disadvantage is that the current price is too high.
- other pies. It is almost the opposite of the Raspberry Pi: it has never lost in terms of hardware parameters, and has never lost in software ecology.
Considering that Radxa Rock is a niche development board, it takes a lot of time just to install the software. After suffering a loss, I only considered the Raspberry Pi this time. It just so happens that my friend has a dusty 3B in his hand, and I plan to take it over and study it first.
The biggest uncertainty now is the microphone and speakers. The initial plan is to buy a second-hand speaker and only use its shell, speaker and microphone; remove the inside and replace it with a Raspberry Pi and related accessories. Try to make a prototype in one or two weeks.
This article is reproduced from: https://atjason.com/daily/2023-07-23.html
This site is only for collection, and the copyright belongs to the original author.