zhouhr / rsnowboy

Snowboy wrapper for rust

rsnowboy

This repo is snowboy wrapper for rust.

Snowboy is created by C++, But rust does not covert all data to C++, So this crate is Snowoby for rust wrapper.

The more document you can see snowboy and kitt.ai

Usage

[dependencies]
rsnowboy = { version = "0.1", git = "https://github.com/fewensa/rsnowboy" }

crates.io

Because crates.io has a limitation

crates.io currently has a 10MB size limit on the .crate file.

So now the new version isn't publish to crates.io. The old version (0.1.0) is also can be use. but not compile itself.

Examples

use rsnowboy::SnowboyDetect;

fn main() {
  let resource_filename = "resources/common.res";
  let model_filename = "resources/models/snowboy.umdl";
  let sensitivity_str = "1";
  let audio_gain = 1 as f32;
  let apply_frontend = false;

  let detector = SnowboyDetect::new(resource_filename, model_filename);

  detector.set_sensitivity(sensitivity_str);
  detector.set_audio_gain(audio_gain);
  detector.apply_frontend(apply_frontend);

  let mut rec_count = 0;
  let mut sil_count = 0;
  let mut save_buffer: Vec<i16> = Vec::new();

  // loop get voice
//  loop {
    let mut voice = voice(); // get voice from your device
    let result = detector.run_short_array_detection(voice.as_ptr(), voice.len() as i32, false);

    match result {
      0 => {
        println!("voice");
        save_buffer.append(voice.as_mut());
        rec_count += 1;
        sil_count = 0;
      }
      1 => {
        println!("Hotword {} detected!", result);
        // play resources/ding.wav
        save_buffer.clear();
        rec_count = 1;
      }
      -2 => {
        println!("silence");
        save_buffer.append(voice.as_mut());
        sil_count += 1;

        if rec_count > 0 {
          save_buffer.append(voice.as_mut());
          sil_count += 1;
          println!("silence");
        }

        if sil_count > 2 && rec_count > 1 {
          /*
          play resources/dong.wav
          There you can write `save_buffer` to wave file, and use any speech to text service convert this voice
          file to text, and do any thing.
          */
          rec_count = 0;
          sil_count = 0;
          save_buffer.clear();
          println!("Listening...");
        }
      }
      _ => {}
    }

//  }

}

fn voice() -> Vec<i16> {
  vec![
    -11, -12, -10, -20, -15, -5, 2, 5, 8,
    10, 12, 14, 13, 20, 25, 14, 12, 9, 12,
    -1, -5, 20, 22, 24, 24, 14, 10, 15, 11
  ]
}

Notice

Because snowboy dependency gsl, so you need install libgsl to your system