Audio Sentence Detector

An advanced audio sentence detection library that uses voice activity detection, silence analysis, and acoustic features to segment audio into sentences.

Installation

npm install audio-sentence-detector

Usage

const AudioSentenceDetector = require('audio-sentence-detector');

// Create detector with custom options
const detector = new AudioSentenceDetector({
    minSilenceDuration: 0.5,
    silenceThreshold: 0.01
});

// Process audio buffer
const sentences = await detector.detect(audioBuffer);

Configuration Options

The AudioSentenceDetector constructor accepts an options object with the following parameters:

Basic Sentence Detection Options

| Option | Default | Description | |--------|---------|-------------| | minSilenceDuration | 0.5 | Minimum duration of silence (in seconds) to be considered a sentence boundary | | silenceThreshold | 0.01 | RMS threshold below which audio is considered silence | | minSentenceLength | 1 | Minimum length of a sentence in seconds | | maxSentenceLength | 15 | Maximum length of a sentence in seconds | | windowSize | 2048 | Size of the analysis window in samples | | idealSentenceLength | 5 | Ideal length of a sentence in seconds (used for probability calculations) | | idealSilenceDuration | 0.8 | Ideal duration of silence between sentences | | allowGaps | true | Whether to allow gaps between sentences | | minSegmentLength | 0 | Minimum length for merged segments | | alignToAudioBoundaries | false | Whether to align sentences with audio file boundaries |

Voice Detection Options

| Option | Default | Description | |--------|---------|-------------| | fundamentalFreqMin | 85 | Minimum fundamental frequency for voice detection (Hz) | | fundamentalFreqMax | 255 | Maximum fundamental frequency for voice detection (Hz) | | voiceActivityThreshold | 0.4 | Threshold for voice activity detection | | minVoiceActivityDuration | 0.1 | Minimum duration of voice activity (seconds) | | energySmoothing | 0.95 | Smoothing factor for energy calculations | | formantEmphasis | 0.7 | Emphasis factor for formant detection | | zeroCrossingRateThreshold | 0.3 | Threshold for zero-crossing rate in voice detection |

Debug Option

| Option | Default | Description | |--------|---------|-------------| | debug | false | Enable debug logging |

Return Value

The detect() method returns an array of sentence objects, each containing:

{
    index: number,          // Index of the sentence
    start: number,          // Start time in seconds
    end: number,           // End time in seconds
    duration: number,      // Duration in seconds
    probability: number    // Confidence score (0-1)
}

Example

const AudioSentenceDetector = require('audio-sentence-detector');

// Create detector with custom settings
const detector = new AudioSentenceDetector({
    minSilenceDuration: 0.3,
    silenceThreshold: 0.02,
    minSentenceLength: 1.5,
    maxSentenceLength: 10,
    debug: true
});

// Process audio file
const fs = require('fs');
const audioBuffer = fs.readFileSync('speech.wav');

try {
    const sentences = await detector.detect(audioBuffer);
    console.log('Detected sentences:', sentences);
} catch (error) {
    console.error('Error processing audio:', error);
}

License

MIT