esp-sr icon indicating copy to clipboard operation
esp-sr copied to clipboard

The program runs normally, but cannot be woken up (AIS-1931)

Open HenTaiCJN opened this issue 11 months ago • 2 comments

Checklist

  • [x] Checked the issue tracker for similar issues to ensure this is not a duplicate.
  • [x] Provided a clear description of your suggestion.
  • [x] Included any relevant context or examples.

Issue or Suggestion Description

I am using esp32s3 and the audio chip is es8388, esp-sr version is 2.0.2.

My feed_channel is four channels. I tried all the "MMNR" combinations, a total of 12 compilations, but none of them worked.

The program can start normally and has the following output

MC Quantized vadnet1:vadnet1_mediumv1_Speech_1_0.5_0.1, min speech:128 ms, min noise:992 ms, mode:0, threshold:0.500, channel:3, tigger:v1 (Feb 18 2025 12:00:55)
MC Quantized wakenet9: WakeNet9_v1h24_小爱同学_3_0.620_0.627, tigger:v3, mode:4, p:0, (Feb 18 2025 12:00:54)
------------detect start------------

But I can't wake up the device no matter how I talk. I printed the i2s_buff in the loop and confirmed that there was data.

i2s_buff[1485] = 4336
i2s_buff[1486] = 25120
i2s_buff[1487] = -6501
i2s_buff[1488] = 30079
i2s_buff[1489] = 14813
i2s_buff[1490] = 7459
i2s_buff[1491] = 29665
i2s_buff[1492] = -10286
i2s_buff[1493] = -1285
i2s_buff[1494] = -27349
i2s_buff[1495] = 11852
i2s_buff[1496] = 10892
i2s_buff[1497] = 1510
i2s_buff[1498] = -19009
i2s_buff[1499] = 10177

HenTaiCJN avatar Mar 12 '25 14:03 HenTaiCJN

I printed out afe_fetch_result_t* res, and the output was the same every time through the loop. I was confused by wake_word_length. Does this variable being 0 mean that the wake word failed to load?

AFE Fetch Result:
  data_size: 1024 bytes
  wakeup_state: 0
  wake_word_index: 0
  vad_state: 0
  trigger_channel_id: 2
  wake_word_length: 0 samples
  ret_value: 0

HenTaiCJN avatar Mar 12 '25 14:03 HenTaiCJN

Hi @HenTaiCJN , The ES8388 does not support 4-channel input; it should only support two microphone inputs. You can try "NM" or "MN". To determine the exact configuration, it is best to export the audio to an SD card for debug.

sun-xiangyu avatar Mar 13 '25 06:03 sun-xiangyu