nx_vad/nx_vad.go

package main

import (
	"flag"
	"nx_vad/dsp/audio"
	"nx_vad/dsp/streaming_vad"
	"fmt"
)

func main () {
    filenamePtr := flag.String("filename", "", "input wav audio file")
    silenceTimePtr := flag.Float64("silence_time", 0.0, "silence time in seconds")
	longestActivatePtr := flag.Float64("longest_activate", 0.0, "longest activate time in seconds")
	flag.Parse()

    var filename string = *filenamePtr
    var silenceTime float32 = float32(*silenceTimePtr)
    var longestActivate float32 = float32(*longestActivatePtr)
	fmt.Println(filename)

	var wavInfo audio.WavInfoType
	err := wavInfo.ParseFromFile(filename)
	if err != nil {
		fmt.Println("wavInfo.ParseFromFile failed.")
	}
	fmt.Printf("sample rate: %d\n", wavInfo.Fmt.SampleRate)

	float32Samples := wavInfo.GetFloat32Samples(0, 2)

	var int16Samples []int16
	var size uint32
	var point int16
	for i := 0; i < len(wavInfo.Data.Sample); i++ {
		point = int16(float32Samples[i] * (1 << 15)) + 1
		int16Samples = append(int16Samples, point)
	}
	size = uint32(len(int16Samples))

	fmt.Printf("sample number: %d\n", size)

	winSize := uint32(7000)
	winStep := uint32(7000)
	count := uint32(0)

	sv := streaming_vad.StreamingVadType{}
	sv.Init(8000, silenceTime, longestActivate)

	var begin uint32
	var end uint32
	//var bufferSize uint32

	for true {
		begin = count * winStep
		end = begin + winSize
		//fmt.Println(end)

		if begin >= size {
			break
		}

		if end >= size {
			end = size
		}
		buffer := int16Samples[begin:end]

		//fmt.Printf("bufferSize: %d\n", bufferSize)
		//fmt.Printf("buffer: %d\n", buffer)

		count++

		err = sv.ProcessSpeechByChunk(buffer)
		if err != nil {
			fmt.Println(err)
			break
		}
		//if count > 3 {
		//	break
		//}
	}

	fmt.Println(len(sv.VadEventMarkerDeque))
	var marker streaming_vad.VadEventMarkerType
	for i := 0; i < len(sv.VadEventMarkerDeque); i++ {
		marker = sv.VadEventMarkerDeque[i]

		fmt.Println(marker.Time)
		fmt.Println(marker.VadFlag)

	}
}