Local Wake

https://github.com/st-matskevich/local-wake
Admin
Local Wake is a lightweight, locally-running wake word detection system for resource-constrained...
Tokens:929
Snippets:11
Trust Score:5.9
License:MIT
Update:1 year ago
Show doc for...
Tokens:
Raw
### Install local-wake using pip

Source: https://github.com/st-matskevich/local-wake/blob/main/README.md

Installs the local-wake package and its dependencies using pip. Ensure Python 3.9+ and pip are installed. This is the primary method for installing the library.

```bash
pip install local-wake
```

--------------------------------

### Install local-wake from Source

Source: https://github.com/st-matskevich/local-wake/blob/main/README.md

Clones the local-wake repository and installs it from source. This method is useful for development or when needing the latest unreleased features.

```bash
git clone https://github.com/st-matskevich/local-wake.git
cd local-wake
pip install .
```

--------------------------------

### Install System Dependencies for PortAudio on Ubuntu

Source: https://github.com/st-matskevich/local-wake/blob/main/README.md

Installs the PortAudio library, a system dependency for the sounddevice Python package, on Ubuntu systems. This is required for audio input functionality.

```bash
sudo apt install libportaudio2
```

--------------------------------

### Run Benchmark Script

Source: https://github.com/st-matskevich/local-wake/blob/main/benchmark/README.md

This command initiates the benchmark evaluation using the default configuration. It allows for performance testing of the local-wake system under various conditions.

```bash
python benchmark.py
```

--------------------------------

### Record Audio Sample using local-wake Library

Source: https://github.com/st-matskevich/local-wake/blob/main/README.md

Records an audio sample to a specified file using the local-wake Python library. Allows control over duration and silence trimming.

```python
import lwake

# Record audio sample
lwake.record("sample.wav", duration=3, trim_silence=True)
```

--------------------------------

### Record Wake Word Reference Sample using CLI

Source: https://github.com/st-matskevich/local-wake/blob/main/README.md

Records an audio sample for a wake word using the local-wake command-line interface. The recording is saved to a specified WAV file. Optional arguments control duration and voice activity detection.

```bash
lwake record ref/sample-1.wav --duration 3 --no-vad
```

--------------------------------

### Perform Real-time Detection using local-wake Library

Source: https://github.com/st-matskevich/local-wake/blob/main/README.md

Initiates real-time wake word detection using the local-wake Python library. It takes a directory of reference files, a threshold, and a callback function to handle detections. The callback receives detection details and the underlying sounddevice stream.

```python
import lwake

# Real-time detection. Callback blocks further listening until return.
# Callback also exposes underlying sounddevice stream if you need to read more audio
def handle_detection(detection, stream):
    print(f"Detected '{detection['wakeword']}' at {detection['timestamp']}")
    # audio, _ = stream.read(16000)                         # Read 1 second of audio
    # soundfile.write("input.wav", audio, samplerate=16000) # Save recording

lwake.listen("reference/folder", threshold=0.1, method="embedding", callback=handle_detection)
```

--------------------------------

### Perform Real-time Wake Word Detection using CLI

Source: https://github.com/st-matskevich/local-wake/blob/main/README.md

Listens for wake words in real-time using the local-wake CLI. It requires a directory of reference .wav files and a detection threshold. Results are printed to stdout in JSON format.

```bash
lwake listen reference/folder 0.1 --method embedding --buffer-size 2.0 --slide-size 0.25 --debug
```

--------------------------------

### Compare Two Audio Files using CLI

Source: https://github.com/st-matskevich/local-wake/blob/main/README.md

Compares two audio files using the local-wake CLI to evaluate similarity. It returns a distance metric that can help determine a suitable detection threshold. Supports 'embedding' or 'mfcc' methods.

```bash
lwake compare ref/sample-1.wav ref/sample-2.wav --method embedding
```

--------------------------------

### Record Wake Word Reference Sample using arecord (Linux)

Source: https://github.com/st-matskevich/local-wake/blob/main/README.md

Records a 3-second audio sample at 16000 Hz, mono, 16-bit LE format using the `arecord` command on Linux. This is an alternative to `lwake record` for creating reference files.

```bash
arecord -d 3 -r 16000 -c 1 -f S16_LE output.wav
```

--------------------------------

### Compare Two Audio Files using local-wake Library

Source: https://github.com/st-matskevich/local-wake/blob/main/README.md

Compares two audio files using the local-wake Python library and returns a distance score. The feature extraction method can be specified.

```python
import lwake

# Compare two audio files
distance = lwake.compare("file1.wav", "file2.wav", method="embedding")
print(f"Distance: {distance}")
```
Tokens:
Raw