88 lines
2.6 KiB
Go
88 lines
2.6 KiB
Go
package main
|
|
|
|
import (
|
|
sherpa "github.com/k2-fsa/sherpa-onnx-go/sherpa_onnx"
|
|
"log"
|
|
)
|
|
|
|
/*
|
|
Usage:
|
|
|
|
Step 1: Download a speaker segmentation model
|
|
|
|
Please visit https://github.com/k2-fsa/sherpa-onnx/releases/tag/speaker-segmentation-models
|
|
for a list of available models. The following is an example
|
|
|
|
wget https://github.com/k2-fsa/sherpa-onnx/releases/download/speaker-segmentation-models/sherpa-onnx-pyannote-segmentation-3-0.tar.bz2
|
|
tar xvf sherpa-onnx-pyannote-segmentation-3-0.tar.bz2
|
|
rm sherpa-onnx-pyannote-segmentation-3-0.tar.bz2
|
|
|
|
Step 2: Download a speaker embedding extractor model
|
|
|
|
Please visit https://github.com/k2-fsa/sherpa-onnx/releases/tag/speaker-recongition-models
|
|
for a list of available models. The following is an example
|
|
|
|
wget https://github.com/k2-fsa/sherpa-onnx/releases/download/speaker-recongition-models/3dspeaker_speech_eres2net_base_sv_zh-cn_3dspeaker_16k.onnx
|
|
|
|
Step 3. Download test wave files
|
|
|
|
Please visit https://github.com/k2-fsa/sherpa-onnx/releases/tag/speaker-segmentation-models
|
|
for a list of available test wave files. The following is an example
|
|
|
|
wget https://github.com/k2-fsa/sherpa-onnx/releases/download/speaker-segmentation-models/0-four-speakers-zh.wav
|
|
|
|
Step 4. Run it
|
|
*/
|
|
|
|
func initSpeakerDiarization() *sherpa.OfflineSpeakerDiarization {
|
|
config := sherpa.OfflineSpeakerDiarizationConfig{}
|
|
|
|
config.Segmentation.Pyannote.Model = "./sherpa-onnx-pyannote-segmentation-3-0/model.onnx"
|
|
config.Embedding.Model = "./3dspeaker_speech_eres2net_base_sv_zh-cn_3dspeaker_16k.onnx"
|
|
|
|
// The test wave file contains 4 speakers, so we use 4 here
|
|
config.Clustering.NumClusters = 4
|
|
|
|
// if you don't know the actual numbers in the wave file,
|
|
// then please don't set NumClusters; you need to use
|
|
//
|
|
// config.Clustering.Threshold = 0.5
|
|
//
|
|
|
|
// A larger Threshold leads to fewer clusters
|
|
// A smaller Threshold leads to more clusters
|
|
|
|
sd := sherpa.NewOfflineSpeakerDiarization(&config)
|
|
return sd
|
|
}
|
|
|
|
func main() {
|
|
wave_filename := "./0-four-speakers-zh.wav"
|
|
wave := sherpa.ReadWave(wave_filename)
|
|
if wave == nil {
|
|
log.Printf("Failed to read %v", wave_filename)
|
|
return
|
|
}
|
|
|
|
sd := initSpeakerDiarization()
|
|
if sd == nil {
|
|
log.Printf("Please check your config")
|
|
return
|
|
}
|
|
|
|
defer sherpa.DeleteOfflineSpeakerDiarization(sd)
|
|
|
|
if wave.SampleRate != sd.SampleRate() {
|
|
log.Printf("Expected sample rate: %v, given: %d\n", sd.SampleRate(), wave.SampleRate)
|
|
return
|
|
}
|
|
|
|
log.Println("Started")
|
|
segments := sd.Process(wave.Samples)
|
|
n := len(segments)
|
|
|
|
for i := 0; i < n; i++ {
|
|
log.Printf("%.3f -- %.3f speaker_%02d\n", segments[i].Start, segments[i].End, segments[i].Speaker)
|
|
}
|
|
}
|