ESP32语音交互之一--------------ESP32-Korvo V1.1开发板初体念

    最近开始开发语音交互,找出了积灰多年的ESP32-LyraT开发板,一顿操作猛如虎,结果板子250,只能唤醒,命令词没搞起来,查看官方文档和GitHub上的代码发现了了一些线索,需要熟悉语音唤醒引擎WakeNet和命令词引擎MultiNet,这两引擎在一个统一的包里,那就是ESP-Skainet,地址:https://github.com/espressif/esp-skainet

   根据文档推荐了两款开发板:ESP32-Korvo V1.1和ESP32-LyraT-Mini,由于有一块双麦的ESP32-LyraT开发板了,就不想体念单麦的ESP32-LyraT-Mini了,选择了230大洋的ESP32-Korvo V1.1,想看看远场交互能力如何。

   关于这款开发板的介绍见网页:https://github.com/espressif/esp-skainet/blob/master/docs/zh_CN/hw-reference/esp32/user-guide-esp32-korvo-v1.1.md,这里面介绍了板子的硬件配置,默认固件说明,PCB原理图,PCB Layout等信息。下面来看下默认固件的体念:

   此默认固件很简单,就是唤醒词“嗨,乐鑫”,然后还有8句指令:

ESP32语音交互之一--------------ESP32-Korvo V1.1开发板初体念

接上电源USB和串口USB线,打开串口终端,上电,按RST键,打印信息如下:

ets Jul 29 2019 12:21:46

rst:0x1 (POWERON_RESET),boot:0x13 (SPI_FAST_FLASH_BOOT)
configsip: 0, SPIWP:0xee
clk_drv:0x00,q_drv:0x00,d_drv:0x00,cs0_drv:0x00,hd_drv:0x00,wp_drv:0x00
mode:DIO, clock div:1
load:0x3fff0018,len:4
load:0x3fff001c,len:7128
load:0x40078000,len:11824
ho 0 tail 12 room 4
load:0x40080400,len:7344
entry 0x40080798
[0;32mI (64) boot: Chip Revision: 3[0m
[0;32mI (71) boot_comm: chip revision: 3, min. bootloader chip revision: 0[0m
[0;32mI (42) boot: ESP-IDF v3.2.3-201-gfda11a6-dirty 2nd stage bootloader[0m
[0;32mI (42) boot: compile time 14:40:42[0m
[0;32mI (42) boot: Enabling RNG early entropy source...[0m
[0;32mI (48) qio_mode: Enabling default flash chip QIO[0m
[0;32mI (53) boot: SPI Speed      : 80MHz[0m
[0;32mI (57) boot: SPI Mode       : QIO[0m
[0;32mI (61) boot: SPI Flash Size : 16MB[0m
[0;32mI (66) boot: Partition Table:[0m
[0;32mI (69) boot: ## Label            Usage          Type ST Offset   Length[0m
[0;32mI (76) boot:  0 factory          factory app      00 00 00010000 003c0000[0m
[0;32mI (84) boot:  1 nvs              WiFi data        01 02 003d0000 00004000[0m
[0;32mI (91) boot: End of partition table[0m
[0;32mI (96) boot_comm: chip revision: 3, min. application chip revision: 0[0m
[0;32mI (103) esp_image: segment 0: paddr=0x00010020 vaddr=0x3f400020 size=0x16fbd4 (1506260) map[0m
[0;32mI (510) esp_image: segment 1: paddr=0x0017fbfc vaddr=0x3ffb0000 size=0x00414 (  1044) load[0m
[0;32mI (511) esp_image: segment 2: paddr=0x00180018 vaddr=0x400d0018 size=0xc7034 (815156) map[0m
[0;32mI (733) esp_image: segment 3: paddr=0x00247054 vaddr=0x3ffb0414 size=0x032bc ( 12988) load[0m
[0;32mI (737) esp_image: segment 4: paddr=0x0024a318 vaddr=0x40080000 size=0x00400 (  1024) load[0m
[0;32mI (740) esp_image: segment 5: paddr=0x0024a720 vaddr=0x40080400 size=0x1959c (103836) load[0m
[0;32mI (797) boot: Loaded app from partition at offset 0x10000[0m
[0;32mI (797) boot: Disabling RNG early entropy source...[0m
WYB:Slave mode:80 !
WYB:ES7210_MCLK_CTL_REG02 = 0xC3!
************** enter es8311 init **************
Quantized wakeNet5: wakeNet5_v1_hilexin_5_0.95_0.90, mode:0 (Nov 20 2019 14:34:43)
Quantized MN1_2 (Dec  5 2019 20:22:00)
SHIFT: 8, 11, 17, 18, 18, 17, 7, 16, 15, 13, 
[0;32mI (287) MN: ---------------------SPEECH COMMANDS---------------------[0m
[0;32mI (297) MN: Command ID0, phrase 0: guan bi dian deng[0m
[0;32mI (297) MN: Command ID1, phrase 1: da kai bai deng[0m
[0;32mI (307) MN: Command ID2, phrase 2: da kai hong deng[0m
[0;32mI (307) MN: Command ID3, phrase 3: da kai lv deng[0m
[0;32mI (317) MN: Command ID4, phrase 4: da kai lan deng[0m
[0;32mI (317) MN: Command ID5, phrase 5: da kai huang deng[0m
[0;32mI (327) MN: Command ID6, phrase 6: da kai cheng deng[0m
[0;32mI (327) MN: Command ID7, phrase 7: da kai zi deng[0m
[0;32mI (337) MN: ---------------------------------------------------------
[0m
ESP32-Korvo V1.1 Firmware V0.1

从上面信息可见这个固件是ESP-IDF v3.2.3-201-gfda11a6-dirty开发环境,比较老的版本了,现在音频基本都是v3.3.2版本开发了。好测试下语言效果:

ESP32语音交互之一--------------ESP32-Korvo V1.1开发板初体念

嗨,乐鑫,打开红灯:

ESP32语音交互之一--------------ESP32-Korvo V1.1开发板初体念

嗨,乐鑫,打开紫灯:

ESP32语音交互之一--------------ESP32-Korvo V1.1开发板初体念

伴随着串口输出:

Wake-up word detected.
phrase ID: 7, prob: 0.951218
1175.010010: command 7 detected.

近距离小声交互都没问题,站在5米元的地方试了下,需要大点声才能交互。总体来说体念还可以!