rustRhino — Rust Quick Start

  • Speech-to-Intent Engine
  • Domain Specific NLU
  • Offline NLU
  • Local Voice Recognition
  • Raspberry Pi
  • BeagleBone
  • NVIDIA Jetson
  • Linux
  • macOS
  • Windows
  • Rust

Requirements

  • Rust 1.54+
  • Cargo

Compatibility

  • Linux (x86_64)
  • macOS (x86_64)
  • Windows (x86_64)
  • Raspberry Pi (all variants)
  • NVIDIA Jetson (Nano)
  • BeagleBone

Cloning the Repository

If using SSH, clone the repository with:

git clone [email protected]:Picovoice/rhino.git

If using HTTPS, then type:

git clone https://github.com/Picovoice/rhino.git

Installation

The microphone demo uses miniaudio-rs for cross-platform audio capture. It uses bindgen and therefore requires clang to be installed and on the path. Use the Bindgen docs for instructions on how to install clang for various Operating Systems and distros.

Usage

NOTE: The working directory for the following Cargo commands is:

rhino/demo/rust/filedemo # File Demo
rhino/demo/rust/micdemo # Microphone Demo

File Demo

The file demo uses Rhino to get an inference result from an audio file. This demo is mainly useful for quantitative performance benchmarking against a corpus of audio data. Note that only the relevant spoken command should be present in the file and no other speech. There also needs to be at least one second of silence at the end of the file.

cargo run --release -- --input_audio_path "path/to/input.wav" \
--context_path "/path/to/context/file.rhn"

The sensitivity of the engine can be tuned using the sensitivity input argument and the default model file can be overridden using the model_path argument:

cargo run --release -- --input_audio_path "path/to/input.wav" \
--context_path "/path/to/context/file.rhn" --sensitivity 0.4

Sensitivity is the parameter that enables trading miss rate for the false alarm rate. It is a floating point number within [0, 1]. A higher sensitivity reduces the miss rate at the cost of increased false alarm rate.

Microphone Demo

The microphone demo opens an audio stream from a microphone and performs inference on spoken commands:

cargo run --release -- --context_path "/path/to/context/file.rhn"

It is possible that the default audio input device is not the one you wish to use. There are a couple of debugging facilities baked into the demo application to solve this. First, type the following into the console:

cargo run --release -- --show_audio_devices

It provides information about various audio input devices on the box. On a is an example output from a Windows machine:

Capture Devices
0: Microphone Array (Realtek(R) Audio)
1: Microphone (USB Audio Device)

You can use the device index to specify which microphone to use for the demo. For instance, if you want to use the USB microphone in the above example, you can invoke the demo application as below:

cargo run --release -- --context_path "/path/to/context/file.rhn" --audio_device_index 1

Exact system setups don't always play well with certain audio backends. If this is the case you can override the default with a specific backend:

cargo run -- --release --context_path "/path/to/context/file.rhn" --audio_device_index 1 --audio_backend Alsa

If the problem persists we suggest storing the recorded audio into a file for inspection. This can be achieved with:

cargo run --release -- --context_path "/path/to/context/file.rhn" --audio_device_index 1 --output_path ./test.wav

If after listening to stored file there is no apparent problem detected please open an issue.

Custom Context

You can create custom Rhino context models using Picovoice Console.


Issue with this doc? Please let us know.