Eagle Binding for Node.js

Eagle Speaker Recognition Engine

Made in Vancouver, Canada by Picovoice

Eagle is an on-device speaker recognition engine. Eagle is:

• Private; All voice processing runs locally.
• Cross-Platform:
  • Linux (x86_64), macOS (x86_64, arm64), Windows (x86_64, arm64)
  • Android and iOS
  • Chrome, Safari, Firefox, and Edge
  • Raspberry Pi (3, 4, 5)

Compatibility

• Node.js 18+
• Runs on Linux (x86_64), macOS (x86_64, arm64), Windows (x86_64, arm64), and Raspberry Pi (3, 4, 5).

Installation

Using yarn:

yarn add @picovoice/eagle-node

or using npm:

npm install --save @picovoice/eagle-node

AccessKey

Eagle requires a valid Picovoice AccessKey at initialization. AccessKey acts as your credentials when using Eagle SDKs. You can get your AccessKey for free. Make sure to keep your AccessKey secret. Signup or Login to Picovoice Console to get your AccessKey.

Usage

Eagle has two distinct steps: Enrollment and Recognition. In the enrollment step, Eagle analyzes a series of utterances from a particular speaker to learn their unique voiceprint. This step produces a profile object, which can be stored and utilized during inference. During the Recognition step, Eagle compares the incoming frames of audio to the voiceprints of all enrolled speakers in real-time to determine the similarity between them.

Speaker Enrollment

Create an instance of the profiler:

const { EagleProfiler } = require("@picovoice/eagle-node");

const accessKey = "${ACCESS_KEY}"; // Obtained from the Picovoice Console (https://console.picovoice.ai/)
const eagleProfiler = new EagleProfiler(accessKey);

EagleProfiler is responsible for processing and enrolling PCM audio data, with the valid audio sample rate determined by eagleProfiler.sampleRate. The audio data must be 16-bit linearly-encoded and single-channel.

When passing samples to eagleProfiler.enroll, the number of samples must be equal to eagleProfiler.frameLength. The percentage value obtained from this process indicates the progress of enrollment.

const { EnrollProgress } = require("@picovoice/eagle-node");

function getAudioData(numSamples): Int16Array {
  // get audio frame of size `numSamples`
}

function hasAudioData(numSamples): Boolean {
  // check if there are any remaining samples in the stream
}

let percentage = 0;
while (percentage < 100 && hasAudioData(eagleProfiler.frameLength)) {
  const audioData = getAudioData(eagleProfiler.frameLength);
  percentage = eagleProfiler.enroll(audioData);
}

percentage = eagleProfiler.flush();

After the percentage reaches 100%, the enrollment process is considered complete. While it is possible to continue providing additional audio data to the profiler to improve the accuracy of the voiceprint, it is not necessary to do so. Once all the audio from a single source has been submitted it is necessary to call flush before submitting any audio from another source.

const speakerProfile: Uint8Array = eagleProfiler.export();

The eagleProfiler.export() function produces a binary array, which can be saved to a file.

To reset the profiler and enroll a new speaker, the eagleProfiler.reset() method can be used. This method clears all previously stored data, making it possible to start a new enrollment session with a different speaker.

Finally, when done be sure to explicitly release the resources:

eagleProfiler.release();

Speaker Recognition

Create an instance of the engine:

const { Eagle } = require("@picovoice/eagle-node");

const accessKey = "${ACCESS_KEY}"; // Obtained from the Picovoice Console (https://console.picovoice.ai/)
const eagle = new Eagle(accessKey);

When initialized, eagle.sampleRate specifies the valid sample rate for Eagle. The minimum length of a sample, or the number of audio samples in an input array, is defined by eagle.minProcessSamples.

Like the profiler, Eagle is designed to work with single-channel audio that is encoded using 16-bit linear PCM.

Process audio with one or more speaker profiles from the EagleProfiler.

function getAudioData(numSamples): Int16Array {
  // get audio frame of size `numSamples`
}

while (true) {
  const audioData = getAudioData(eagle.minProcessSamples);
  const scores: number[] = eagle.process(audioData, speakerProfile);
}

The scores will be null or be an array that contains floating-point numbers that indicate the similarity between the input audio frame and the enrolled speakers. Each value in the array corresponds to a specific enrolled speaker, maintaining the same order as the speaker profiles provided during initialization. The values in the array range from 0.0 to 1.0, where higher values indicate a stronger degree of similarity. A result of null indicates that there was not enough voice in the audio to recognize any speakers.

Finally, when done be sure to explicitly release the resources:

eagle.release();

Demos

The Eagle Node.js demo package provides command-line utilities for processing audio using Eagle.