Put VAD back as the IVR platform needs it to stop audio when input is detected

[MayamaTakeshi]

VAD was removed here due to problems with node-vad:
e6c98b1

Before trying again with node-vad, lets try this one:
https://github.com/OzymandiasTheGreat/libfvad-wasm

I did some experiments and it looks good:

import wav from 'wav'
import fs from 'fs'

import  Speaker from 'speaker'

import VADBuilder, { VADMode } from "@ozymandiasthegreat/vad";


const reader = new wav.Reader()

if (process.argv.length != 4) {
console.log("Expected: file_path output_file")
process.exit(1)
}
const file_path = process.argv[2]
const output_file = process.argv[3]

const file = fs.createReadStream(file_path)

file.pipe(reader)

reader.on('format', format => {
console.log('format', format)

VADBuilder().then((VAD) => {
console.log('VAD ready')
const vad = new VAD(VADMode.VERY_AGGRESSIVE, format.sampleRate);

reader.on('data', data => {
 //console.log('data.byteLength', data.byteLength)
 const size = 640
 var i = 0
 var chunk = data.buffer.slice(i, size)
 while(chunk.byteLength) {
           //console.log(chunk)
 //console.log('chunk.byteLength', chunk.byteLength)
 var samples = new Int16Array(chunk)
 var res = vad.processBuffer(samples)
     var output = Array.from(samples).map(sample => `${sample},${res}`);
 console.log(output)
 fs.appendFileSync(output_file, output.join('\n') + '\n');

 i += size
         chunk = data.buffer.slice(i, i+size)
         }
})

const speaker = new Speaker(format)
reader.pipe(speaker)
});
})

The above will generate a CSV file with samples and VAD result for each one of them.
This can be visualized using matplotlib:

import numpy as np
import matplotlib.pyplot as plt

# Read the output file
data = np.genfromtxt('res.csv', delimiter=',')

# Separate audio samples and VAD results
audio_samples = data[:, 0]
vad_results = data[:, 1]

# Plot audio signal
plt.figure(figsize=(10, 5))
plt.plot(audio_samples, label='Audio Signal')
plt.xlabel('Sample')
plt.ylabel('Amplitude')

# Overlay VAD results
vad_indices = np.where(vad_results == 1)[0]
plt.scatter(vad_indices, audio_samples[vad_indices], color='red', label='Voice Activity')

plt.legend()
plt.title('Audio Signal with VAD')
plt.show()