voicebot/stt/sttselvas.go

//for STT SELVAS STT
package stt

/*
#cgo LDFLAGS: -lstdc++ -lssl -lcrypto /home/leejj9612/dev/voiceagent/extlib/selvasstt/SDK/LIB/c_linux/x64/libASRLIB.a
#cgo CFLAGS: -I /home/leejj9612/dev/voiceagent/extlib/selvasstt/SDK/INCLUDE
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <Lvcsr_Api.h>

char* getResultData(LVCSR_DATA_RESULT* pDataResult, long long nCount) {
	char* skip1 = "<eps>";
	char* space = " ";
	int len = 0;
	long long i;

	for (i = 0; i < nCount; i++) {
		len += strlen(pDataResult[i].pTokenStr) + 1;
	}

	char* result = malloc(sizeof(char) * len);
	strcpy(result, "");

	for (i = 0; i < nCount; i++) {
		if (strcmp(pDataResult[i].pTokenStr,skip1) == 0) {
		} else {
			strcat(result, space);
			strcat(result, pDataResult[i].pTokenStr);
		}
	}

	return result;
}

*/
import "C"
import (
	"fmt"
	"os"
	"strings"
	"sync"
	"unsafe"

	"gitlab.com/cinnamon/voiceagent/icserror"
)

const (
	CONNECT_TIMEOUT      = 3
	READ_CONNECT_TIMEOUT = 5

	// model Info
	MODEL_ID   = 0
	KWD_ID     = -1
	CODEC_TYPE = 0 // 8k
	LANGUAGE   = 1 // utf-8
	USED_EPD   = 1 // epd used
	USED_SCORE = 0 // used off
)

type STTSelvas struct {
	handle       int
	authCode     string
	ch           int
	text         string
	voiceBuf     []byte
	voiceBufCur  int64
	silencenum   int
	validnum     int64 //rms counter
	uDataSize    int
	uEndOfSpeech int
	STTInfo      STTInfo
}

type STTInfo struct {
	LVCSR_SOCK_HEAD           C.LVCSR_SOCK_HEAD
	LVCSR_EPD_INFO            C.LVCSR_EPD_INFO
	LVCSR_DATA_AUTHENTICATION C.LVCSR_DATA_AUTHENTICATION
	LVCSR_RECOG_RESULT        C.LVCSR_RECOG_RESULT
	LVCSR_DATA_RESULT         C.LVCSR_DATA_RESULT
	LVCSR_RECOG_MID_RESULT    C.LVCSR_RECOG_MID_RESULT
	LVCSR_DATA_INFO           C.LVCSR_DATA_INFO
}

type STTSResult struct {
	result string
	error  *icserror.IcsError
}

// connect SELVAS STT Server
func NewSTTS(IP string, port int) (*STTSelvas, *icserror.IcsError) {
	if len(IP) <= 0 || port <= 0 {
		return nil, icserror.ICSERRInvalidParam
	}
	// os.Exit(9)

	stts := STTSelvas{handle: -1, authCode: "LGUPlusManager", uDataSize: 1600, uEndOfSpeech: 0}

	csIP := C.CString(IP)
	csPort := C.long(port)
	csConTimeout := C.long(CONNECT_TIMEOUT)
	csReadTimeout := C.long(READ_CONNECT_TIMEOUT)

	defer func() {
		C.free(unsafe.Pointer(csIP))
	}()

	rc := C.ASR_SVC_OPEN(csIP, csPort, csConTimeout, csReadTimeout, &stts.STTInfo.LVCSR_SOCK_HEAD) // add LVCSR_SOCK_HEAD
	if int(rc) == -1 {
		return nil, icserror.ICSERRSTTConnectTimeout
	} else if int(rc) == -2 {
		return nil, icserror.ICSERRSTTFailInit
	}

	// Auth
	csAuthCode := C.CString(stts.authCode)
	csAuthLen := C.longlong(len(stts.authCode))
	stts.STTInfo.LVCSR_DATA_AUTHENTICATION.nAuthenticationLen = csAuthLen
	stts.STTInfo.LVCSR_DATA_AUTHENTICATION.pAuthentication = csAuthCode
	defer C.free(unsafe.Pointer(csAuthCode))

	rc = C.ASR_SVC_SET_AUTH(&stts.STTInfo.LVCSR_SOCK_HEAD, &stts.STTInfo.LVCSR_DATA_AUTHENTICATION)
	if int(rc) < 0 {
		return nil, icserror.ICSERRSTTFailInit
	}

	// Channel Connect
	rc = C.ASR_SVC_RECG_OPEN(&stts.STTInfo.LVCSR_SOCK_HEAD) // add LVCSR_SOCK_HEAD
	if int(rc) < 0 {
		return nil, icserror.ICSERRSTTFailInit
	}

	return &stts, nil
}

func (s *STTSelvas) Close() *icserror.IcsError {
	if s.handle < 0 || s.ch < 0 {
		return icserror.ICSERRSTTNotInit
	}

	// Channel Connection Close
	rc := C.ASR_SVC_RECG_CLOS(&s.STTInfo.LVCSR_SOCK_HEAD) // add LVCSR_SOCK_HEAD
	if int(rc) < 0 {
		return icserror.ICSERRTTSNotInit
	}

	// Server Close
	rc = C.ASR_SVC_CLOS(&s.STTInfo.LVCSR_SOCK_HEAD) // aadd LVCSR_SOCK_HEAD
	if int(rc) < 0 {
		return icserror.ICSERRTTSNotInit
	}
	return nil
}

func (s *STTSelvas) SendSTT(voicedata []byte) (string, *icserror.IcsError) {
	var result string
	var sendCount int

	// file open
	// buff := make([]byte, 1600)

	file, err := os.Open("./9001-RX-1648533187911907029.pcm")
	if err != nil {
		fmt.Println(err)
	}
	defer file.Close()

	// Search model list
	/*
		rc := C.ASR_SVC_RECG_LIST_VIEW(&s.STTInfo.LVCSR_SOCK_HEAD, &s.STTInfo.LVCSR_DATA_MODEL)
		if int(rc) < 0 {
			return icserror.ICSERRSTTFailInit
		}
		fmt.Printf("Model List : %d \n", s.STTInfo.LVCSR_DATA_MODEL.nModelCnt)
		fmt.Println("Model List : ", s.STTInfo.LVCSR_DATA_MODEL.pModelInfo)
		fmt.Println("Model List : ", s.STTInfo.LVCSR_DATA_MODEL.pModelInfo.nModelID)
		fmt.Println("Model List : ", *s.STTInfo.LVCSR_DATA_MODEL.pModelInfo.pModelName)
		fmt.Println("Model List : ", s.STTInfo.LVCSR_DATA_MODEL.pModelInfo.nModelType)
		fmt.Println("Model List : ", s.STTInfo.LVCSR_DATA_MODEL.pModelInfo.nSamplingRate)
		// for i := 0; i < s.STTInfo.LVCSR_DATA_MODEL.nModelCnt; i++ {
		// 	fmt.Println("Model List : ", s.STTInfo.LVCSR_DATA_MODEL.pModelInfo[i])
		// }

		rc = C.ASR_SVC_RECG_LIST_VIEW_FREE(&s.STTInfo.LVCSR_DATA_MODEL)
		if int(rc) < 0 {
			return icserror.ICSERRSTTFailInit
		}
	*/

	// Set Model List
	s.STTInfo.LVCSR_DATA_INFO.nModelId = MODEL_ID
	s.STTInfo.LVCSR_DATA_INFO.nKwdId = KWD_ID
	s.STTInfo.LVCSR_DATA_INFO.nCodecType = CODEC_TYPE
	s.STTInfo.LVCSR_DATA_INFO.nCharSet = LANGUAGE
	s.STTInfo.LVCSR_DATA_INFO.bEpdUsed = USED_EPD
	s.STTInfo.LVCSR_DATA_INFO.bScoreUsed = USED_SCORE

	rc := C.ASR_SVC_RECG_SET_LIST(&s.STTInfo.LVCSR_SOCK_HEAD, &s.STTInfo.LVCSR_DATA_INFO)
	if int(rc) < 0 {
		return "", icserror.ICSERRSTTFailInit
	}

	// file split
	sendCount = len(voicedata) / 1600

	// send voice(voicedata)
	for j := 0; j < sendCount; j++ {
		buff1 := voicedata[1600*j : 1600*(j+1)]
		csUDataSize := C.long(s.uDataSize)
		csUEndSpeech := C.long(s.uEndOfSpeech)
		csBuff := (*C.char)(unsafe.Pointer(&buff1[0]))
		rc := C.ASR_SVC_RECG_DATA(&s.STTInfo.LVCSR_SOCK_HEAD, csBuff, csUDataSize, csUEndSpeech, &s.STTInfo.LVCSR_EPD_INFO)
		if int(rc) < 0 {
			return "", icserror.ICSERRSTTSendFail
		}

		if s.STTInfo.LVCSR_EPD_INFO == 2 {
			break
		}
		// time.Sleep(time.Millisecond * 100) // for real time
	}

	// Send voice
	// for {
	// 	bytesRead, err := file.Read(buff)
	// 	if err != nil {
	// 		if err == io.EOF {
	// 			s.uEndOfSpeech = 1
	// 			csUEndSpeech := C.long(s.uEndOfSpeech)
	// 			rc = C.ASR_SVC_RECG_DATA(&s.STTInfo.LVCSR_SOCK_HEAD, nil, 0, csUEndSpeech, &s.STTInfo.LVCSR_EPD_INFO)
	// 			if int(rc) < 0 {
	// 				return "", icserror.ICSERRSTTSendFail
	// 			}
	// 			break
	// 		}
	// 		break
	// 	}
	// 	if bytesRead <= 0 {
	// 		s.uDataSize = 0
	// 	}
	// 	csUDataSize := C.long(s.uDataSize)
	// 	csUEndSpeech := C.long(s.uEndOfSpeech)
	// 	csBuff := (*C.char)(unsafe.Pointer(&buff[0]))

	// 	rc := C.ASR_SVC_RECG_DATA(&s.STTInfo.LVCSR_SOCK_HEAD, csBuff, csUDataSize, csUEndSpeech, &s.STTInfo.LVCSR_EPD_INFO)
	// 	if int(rc) < 0 {
	// 		return "", icserror.ICSERRSTTSendFail
	// 	}

	// 	if s.STTInfo.LVCSR_EPD_INFO == 2 {
	// 		break
	// 	}

	// 	time.Sleep(time.Millisecond * 100) // for real time
	// }

	if s.STTInfo.LVCSR_EPD_INFO != 2 {
		s.uEndOfSpeech = 1
		csUEndSpeech := C.long(s.uEndOfSpeech)

		rc = C.ASR_SVC_RECG_DATA(&s.STTInfo.LVCSR_SOCK_HEAD, nil, 0, csUEndSpeech, &s.STTInfo.LVCSR_EPD_INFO)
		if int(rc) < 0 {
			return "", icserror.ICSERRSTTSendFail
		}
	}

	rc = C.ASR_SVC_RECG_PROC(&s.STTInfo.LVCSR_SOCK_HEAD, &s.STTInfo.LVCSR_RECOG_RESULT)
	if int(rc) < 0 {
		return "", icserror.ICSERRSTTSendFail
	}

	wg := sync.WaitGroup{}
	wg.Add(1)
	go func() {
		result = C.GoString(C.getResultData(s.STTInfo.LVCSR_RECOG_RESULT.pDataResult, s.STTInfo.LVCSR_RECOG_RESULT.nCount))
		// result = C.GoString(C.getResultData1(s.STTInfo.LVCSR_RECOG_RESULT.pDataResult, s.STTInfo.LVCSR_RECOG_RESULT.nCount))
		strings.TrimLeft(result, "")
		wg.Done()
	}()
	wg.Wait()

	// recog memory free
	rc = C.ASR_SVC_RECG_PROC_FREE(&s.STTInfo.LVCSR_RECOG_RESULT)
	if int(rc) < 0 {
		return "", icserror.ICSERRSTTSendFail
	}

	return result, icserror.ICSERRSTTContinue
}

// result return
func NewSTTSResult(result string, err *icserror.IcsError) *STTSResult {
	return &STTSResult{result, err}
}