39 lines
1.7 KiB
Markdown
39 lines
1.7 KiB
Markdown
# Generating Audio with Tensorflow Liete
|
|
|
|
The starting point is the good overview provided by [the "Hallo World" example of Tensorflow Lite](https://www.tensorflow.org/lite/microcontrollers/get_started_low_level#train_a_model) which describes how to create, train and use a model which based on the __sine function__.
|
|
|
|
We just send a generated sine wave to the I2S interface. Further information can be found in the [Wiki](https://github.com/pschatzmann/arduino-audio-tools/wiki/Tensorflow-Lite----Audio-Output)
|
|
Please note the log level should be set so that there is no disturbing output!
|
|
|
|
|
|
### External DAC:
|
|
|
|
For my tests I am using the 24-bit PCM5102 PCM5102A Stereo DAC Digital-to-analog Converter PLL Voice Module pHAT
|
|
|
|
|
|
|
|
I am just using the default pins defined by the framework. However I could change them with the help of the config object. The mute pin can be defined in the constructor of the I2SStream - by not defining anything we use the default which is GPIO23
|
|
|
|
|
|
DAC | ESP32
|
|
-----|----------------
|
|
VCC | 5V
|
|
GND | GND
|
|
BCK | BCK (GPIO14)
|
|
DIN | OUT (GPIO22)
|
|
LCK | BCK (GPIO15)
|
|
FMT | GND
|
|
XMT | 3V (or another GPIO PIN which is set to high)
|
|
|
|
- DMP - De-emphasis control for 44.1kHz sampling rate(1): Off (Low) / On (High)
|
|
- FLT - Filter select : Normal latency (Low) / Low latency (High)
|
|
- SCL - System clock input (probably SCL on your board).
|
|
- FMT - Audio format selection : I2S (Low) / Left justified (High)
|
|
- XMT - Soft mute control(1): Soft mute (Low) / soft un-mute (High)
|
|
|
|
## Dependencies
|
|
|
|
You need to install the following libraries:
|
|
|
|
- https://github.com/pschatzmann/arduino-audio-tools
|
|
- https://github.com/pschatzmann/tflite-micro-arduino-examples |