Recently, the voice recognition of Dajiang North, must be the area of major companies to compete, and major platforms are in their own speech recognition cloud platform, and many application vendors want to quickly use and have a comparison of cloud platforms. Large effort, spending a lot of human and time, often hard to get better results. Do you want to play the voice recognition in just a few days, then modify the modified code to easily productize it? Then you need to know about Respeaker Core V2.0.
Respeaker Core V2.0 Intelligent Voice Development Board Purchase Link. 1882.
Respeaker Core V2.0 Introduction
Respeaker Core V2.0 is designed for voice interface applications. It is based on the RockChip RK3229 of the quad-core ARM Cortex A7 with a running frequency of 1.5GHz, with 1GB RAM. The circuit board has six microphone arrays, and the speech algorithm includes DOA (reaching direction), BF (beamforming), AEC (acoustic echo cancellation), and the like. Run the GNU / Linux operating system in Respeaker Core V2.0. Benefiting from a powerful and active community, the development board allows development, testing, and deployment of existing software and tools to achieve rapid product development. The original intention of the Respeaker Core V2.0 is to meet the diverse business needs. The development plate is mainly divided into two parts, and the first part is to include CPU, memory (RAM), and PMU. The second part is an external carrier containing peripherals, such as EMMC, connector, and wireless connection components. The two parts of the development board can be customized through the service of SEEED.
Respeaker Core V2.0 out of the box
Ok, let's experience it together.
The first day of the first day of just got was smashed by the owner of the cat in the home, and it was found that the taste of the box was very good.
The front package is printed with the full English instructions, and the respeaker core v2 is briefly introduced.
The back of the outer packaging is the introduction of SEEED and the list of equipment.
After opening the package, you can see that the core processor board is attached to the bottom plate through post hole, and the hexagonal plate is very special, randomly with a MicroSB data cable, and does not give paper-related information and instructions. (Wiki URL) can be found on the official website).
Respeaker Core V2.0 has a large change in V1.0, first of all, the original two spliced boards becomes now integrated on a board. The processor is also upgraded by the original MT7688 to the RK3229 of the processing function. The microphone is upgraded from the original microphone to the current 6 microphone arrays. These changes have made V2 version of Respeaker Core more excellent performance.
The specification parameters of the entire board are as follows:
Respeaker Core V2.0 Frontboard resources are as follows:
The back distribution is arranged with a microphone and RGB LED lights and audio amplification drive chips.
The entire system architecture is as follows:
The definition of the extended needle is as follows:
Master chip rk3229
The main control is used by Ruixin micro RK3229. The middle end of performance positioning is mainly used for IPTV / OTT set-top box products. It features 28 nano-processes, with quad-core Cortex-A7, the highest frequency is up to 1.5GHz, support 4K 10bit H.264 /H.265 decoded, support 4K 60FPS TV display.
Aiming to speech recognition of Ruixin is also a plan to give 6 wheat arrays, and the official program block diagram is as follows:
At the same time, the corresponding sound source direction, noise suppression, beamforming, echo cancellation, reverberation, long-distance pickup, model matching microphone array algorithm.
High performance four-channel data switcher ADC AC108
The AC108 is a high-performance four-channel data converter ADC, mainly to smart voice far field microphone array pickup, up to level 16-channel microphone. The AC108's SNR is 108dB, which is the current industry. Therefore, after the product is launched, it has obtained major voice engine companies, algorithm companies, and quickly mass production in smart speakers, TV, OTT boxes and USB peripherals.
The Respeaker Core V2.0 uses 2 AC108 to acquire, process and data conversion of the analog signals of 6 microphones.
Microphone, RGB LED light
The development board uses a patch simulated microphone, silk prints the words S1963 2892, distributed on 6 feet of hexagonal board.
The LED driver uses 3 triode string-limit current resistors, 8 RGB LEDs use a series of in-line, evenly distributed in the bottom plate.
WiFi, Bluetooth, FM module AP6212
WiFi, Bluetooth AP6212, this is a combination of WiFi + BT + FM. The module combines the IEEE 802.11 / b / g / n standard, and the module provides the SDIO interface for WiFi, providing the UART / PCM interface for Bluetooth, while FM provides a UART / I2C / PCM interface.
Mono audio amplifier S8508E
The CS8508E is a high efficiency, ultra-low EMI8.0W mono audio amplifier. In the case where the power supply voltage is 7.4V, the CS8508E can output 6.8 W of the load output of 4 Euro. The PWM modulation of CS8508E does not require the filter reduces external components, PCB area and system cost, and also simplifies the design. 2.5 ~ 8.8V wide voltage work range, the D-class mode up to 90% efficiency, the AB class D mode can be switchable Ω load output 6.8W power. Fast startup time and fiber package size in D-mode make CS8508E a two-segment lithium battery in the case of the most suitable audio amplifier in series. Onboard CS8508E provides a drive for the JST2.0 audio output interface.
Used
Use the instruction manual of Respeaker Core V2.0 on the official website of SEEED: Web address.
Before you start using you, you need to prepare a TF card greater than 4G and a sound. Small built-in Xiaomi Bluetooth speaker weld the horn to the jst2.0 audio interface.
According to Wiki instructions, if you need to use Respeaker Core V2.0, it is recommended to use the officially provided LXQT + SD version of the system image file.
However, Xiaobian downloaded the image of the download link provided by Wiki in the process of using the speech recognition example, the MRAA error has been reported, and then the official updated Wiki but when the SHA256 check is not consistent, it cannot burn, so Xiaobian Seriously read the official post on Respeaker Core V2.0, I finally found the correct system mirror "Respeaker-Debian-9-LXQT-SD-20180610-4Gb.img.xz". Wasted for a long time, I hope the official will mention the document to download the test.
Give the correct firmware download link here: Download address
ObT can use etcher, open the software on the computer, insert the TF card, follow the prompt step to burn the system firmware, after the burn is completed, insert the TF card into the development board, the User1 of the front of the board is on the front of the board, and the USER2 light will flash back and forth until there is a lamp Changliang is complete of the system installation, and the installation process takes about 10 minutes. The entire system installation process is similar to the Raspberry Pie, which is relatively simple.
After completing the system installation, use the Mircro USB to connect to the OTG interface. After the system starts, a serial device appears on the computer, and Windows 10 can search the installation driver.
The serial port can be laulinated by the serial port of PUTTY.
The next step can be performed according to the instruction document prompts and after the system is updated.
Respeaker Core V2.0 provides an example of speech recognition app, which is a small presenter to demonstrate ALEXA AMAZON Alexa.
Step 1: Install the AVS library (Python)
Sudo Apt Update
PIP Install AVS
Step 2: Authorization Alexa
Open the terminal input instruction via VNC.
~ / .local / bin / Alexa-Auth
The desktop automatically pops up the landing page, and log in to the Amazon account to complete the authorization.
Step 3: Install alexa app
CD ~
Git clone https://github.com/voice-ENGINE/voice-engine.git
CD ~ / voice-engine / example
Step 4: Run alexa app
Python kWS_DOA_ALEXA_RESPEAKER_V2.PY
This example contains the DOA sound source directional detection. From different angles, the wake-up word "Alexa" gets different angular values, as shown in the figure, the first and last wake-up wake up in the same orientation.
Below is the actual effect of AVS, the process requires scientific Internet access, the effect is still very good, the actual waken distance can also reach the manual for more than 5 meters, and it is also possible to watch the World Cup.
Test video
summary
Today, the voice recognition service platform is getting more and more accurate, the identification is increasing, a high integration, easy development, strong function, cost-effective speech recognition hardware platform will be more popular, Respeaker Core V2.0 is this hardware, saving a lot of development costs, allowing the product to get online, which is what many engineers and intelligent hardware manufacturers are expected.
Overall experience, hardware platform performance is good and very good. It is also feeling that it is very no need to have a problem with the link firmware in the documentation, and the response speed of the forum is still not enough, and it is hoped that the official can correct it in a timely manner.
It is worth mentioning that by writing Python code, you can use MRAA to operate some peripherals to yourself, you can use the respeaker as a voice controller, such as by connecting the wireless gateway through the onboard extension data interface. The smart device of the gateway is controlled, is it a cool thing? Interested or take it.
you may also like:
2018 Domestic Engineer / Chuangguan is the most worthy of development board TOP 10 ranking
Our other product:
