请帮助我解决这个问题我差不多1周解决这个但我不能请帮助我这是我的日志结果是相同的演示这是我在RecognizeOptions中的代码
private RecognizeOptions getRecognizeOptions(InputStream captureStream) {
return new RecognizeOptions.Builder()
.timestamps(true)
.contentType(HttpMediaType.AUDIO_MP3)
.interimResults(true)
.smartFormatting(true)
.inactivityTimeout(2000)
.speakerLabels(true)
.audio(captureStream)
.wordAlternativesThreshold((float) 0.01)
.wordConfidence(true)
.model("en-US_BroadbandModel")
.build();
}
这是显示说话者和单词的代码
私有类MicrophoneRecognizeDelegate扩展BaseRecognizeCallback {
@Override
public void onTranscription(SpeechRecognitionResults speechResults) {
System.out.println(speechResults);
SpeakerLabelsDiarization.RecoTokens recoTokens = new SpeakerLabelsDiarization.RecoTokens();
if(speechResults.getSpeakerLabels() !=null) {
recoTokens.add(speechResults);
String speaks = "Speaker " + speechResults.getSpeakerLabels().get(0).getSpeaker().toString();
showMicText(speaks);
System.out.println(speaks);
}if(speechResults.getResults() != null && !speechResults.getResults().isEmpty()) {
String text = speechResults.getResults().get(0).getAlternatives().get(0).getTranscript();
showMicText1(text);
}
}
关于SpeakerLabelsDiarization.RecoTokens recoTokens = new SpeakerLabelsDiarization.RecoTokens（）;这是我在github的WatBot示例中获取此代码的代码
package com.example.ezminute.activities;
import com.ibm.watson.developer_cloud.speech_to_text.v1.model.RecognizeOptions;
import com.ibm.watson.developer_cloud.speech_to_text.v1.model.SpeakerLabelsResult;
import com.ibm.watson.developer_cloud.speech_to_text.v1.model.SpeechRecognitionAlternative;
import com.ibm.watson.developer_cloud.speech_to_text.v1.model.SpeechRecognitionResult;
import com.ibm.watson.developer_cloud.speech_to_text.v1.model.SpeechRecognitionResults;
import com.ibm.watson.developer_cloud.speech_to_text.v1.model.SpeechTimestamp;
import com.ibm.watson.developer_cloud.speech_to_text.v1.websocket.BaseRecognizeCallback;
import com.ibm.watson.developer_cloud.util.GsonSingleton;
import java.util.ArrayList;
import java.util.LinkedHashMap;
import java.util.List;
import java.util.Map;
import java.util.Set;
import java.util.concurrent.CountDownLatch;
public class SpeakerLabelsDiarization {
public static class RecoToken {
private Double startTime;
private Double endTime;
private Long speaker;
private String word;
private Boolean spLabelIsFinal;
/**
* Instantiates a new reco token.
*
* @param speechTimestamp the speech timestamp
*/
RecoToken(SpeechTimestamp speechTimestamp) {
startTime = speechTimestamp.getStartTime();
endTime = speechTimestamp.getEndTime();
word = speechTimestamp.getWord();
}
/**
* Instantiates a new reco token.
*
* @param speakerLabel the speaker label
*/
RecoToken(SpeakerLabelsResult speakerLabel) {
startTime = Double.valueOf(speakerLabel.getFrom());
endTime = Double.valueOf(speakerLabel.getTo());
speaker = speakerLabel.getSpeaker();
}
/**
* Update from.
*
* @param speechTimestamp the speech timestamp
*/
public void updateFrom(SpeechTimestamp speechTimestamp) {
word = speechTimestamp.getWord();
}
/**
* Update from.
*
* @param speakerLabel the speaker label
*/
public void updateFrom(SpeakerLabelsResult speakerLabel) {
speaker = speakerLabel.getSpeaker();
}
}
/**
* The Class Utterance.
*/
public static class Utterance {
private Integer speaker;
private String transcript;
/**
* Instantiates a new utterance.
*
* @param speaker the speaker
* @param transcript the transcript
*/
public Utterance(final Integer speaker, final String transcript) {
this.speaker = speaker;
this.transcript = transcript;
}
}
/**
* The Class RecoTokens.
*/
public static class RecoTokens {
private Map<Double, RecoToken> recoTokenMap;
/**
* Instantiates a new reco tokens.
*/
public RecoTokens() {
recoTokenMap = new LinkedHashMap<Double, RecoToken>();
}
/**
* Adds the.
*
* @param speechResults the speech results
*/
public void add(SpeechRecognitionResults speechResults) {
if (speechResults.getResults() != null)
for (int i = 0; i < speechResults.getResults().size(); i++) {
SpeechRecognitionResult transcript = speechResults.getResults().get(i);
if (transcript.isFinalResults()) {
SpeechRecognitionAlternative speechAlternative = transcript.getAlternatives().get(0);
for (int ts = 0; ts < speechAlternative.getTimestamps().size(); ts++) {
SpeechTimestamp speechTimestamp = speechAlternative.getTimestamps().get(ts);
add(speechTimestamp);
}
}
}
if (speechResults.getSpeakerLabels() != null)
{
for (int i = 0; i < speechResults.getSpeakerLabels().size(); i++) {
add(speechResults.getSpeakerLabels().get(i));
}
}
}
/**
* Adds the.
*
* @param speechTimestamp the speech timestamp
*/
public void add(SpeechTimestamp speechTimestamp) {
RecoToken recoToken = recoTokenMap.get(speechTimestamp.getStartTime());
if (recoToken == null) {
recoToken = new RecoToken(speechTimestamp);
recoTokenMap.put(speechTimestamp.getStartTime(), recoToken);
} else {
recoToken.updateFrom(speechTimestamp);
}
}
/**
* Adds the.
*
* @param speakerLabel the speaker label
*/
public void add(SpeakerLabelsResult speakerLabel) {
RecoToken recoToken = recoTokenMap.get(speakerLabel.getFrom());
if (recoToken == null) {
recoToken = new RecoToken(speakerLabel);
recoTokenMap.put(Double.valueOf(speakerLabel.getFrom()), recoToken);
} else {
recoToken.updateFrom(speakerLabel);
}
if (speakerLabel.isFinalResults()) {
markTokensBeforeAsFinal(speakerLabel.getFrom());
report();
cleanFinal();
}
}
private void markTokensBeforeAsFinal(Float from) {
Map<Double, RecoToken> recoTokenMap = new LinkedHashMap<>();
for (RecoToken rt : recoTokenMap.values()) {
if (rt.startTime <= from)
{
rt.spLabelIsFinal = true;
}
}
}
/**
* Report.
*/
public void report() {
List<Utterance> uttterances = new ArrayList<Utterance>();
Utterance currentUtterance = new Utterance(0, "");
for (RecoToken rt : recoTokenMap.values()) {
if (currentUtterance.speaker != Math.toIntExact(rt.speaker)) {
uttterances.add(currentUtterance);
currentUtterance = new Utterance(Math.toIntExact(rt.speaker), "");
}
currentUtterance.transcript = String.format("%s%s ", currentUtterance.transcript, rt.word);
}
uttterances.add(currentUtterance);
String result = GsonSingleton.getGson().toJson(uttterances);
System.out.println(result);
}
private void cleanFinal() {
Set<Map.Entry<Double, RecoToken>> set = recoTokenMap.entrySet();
for (Map.Entry<Double, RecoToken> e : set) {
if (e.getValue().spLabelIsFinal) {
recoTokenMap.remove(e.getKey());
}
}
}
}
private static CountDownLatch lock = new CountDownLatch(1);
}
这是我的结果
{
2019-03-03 23:21:58.335 27985-28793/com.example.ezminute I/System.out: "speaker_labels": [
2019-03-03 23:21:58.335 27985-28793/com.example.ezminute I/System.out: {
2019-03-03 23:21:58.335 27985-28793/com.example.ezminute I/System.out: "confidence": 0.605,
2019-03-03 23:21:58.335 27985-28793/com.example.ezminute I/System.out: "final": false,
2019-03-03 23:21:58.335 27985-28793/com.example.ezminute I/System.out: "from": 2.43,
2019-03-03 23:21:58.335 27985-28793/com.example.ezminute I/System.out: "speaker": 0,
2019-03-03 23:21:58.335 27985-28793/com.example.ezminute I/System.out: "to": 2.93
2019-03-03 23:21:58.335 27985-28793/com.example.ezminute I/System.out: },
2019-03-03 23:21:58.335 27985-28793/com.example.ezminute I/System.out: {
2019-03-03 23:21:58.335 27985-28793/com.example.ezminute I/System.out: "confidence": 0.667,
2019-03-03 23:21:58.335 27985-28793/com.example.ezminute I/System.out: "final": false,
2019-03-03 23:21:58.335 27985-28793/com.example.ezminute I/System.out: "from": 3.3,
2019-03-03 23:21:58.336 27985-28793/com.example.ezminute I/System.out: "speaker": 1,
2019-03-03 23:21:58.336 27985-28793/com.example.ezminute I/System.out: "to": 3.82
2019-03-03 23:21:58.336 27985-28793/com.example.ezminute I/System.out: },
2019-03-03 23:21:58.336 27985-28793/com.example.ezminute I/System.out: {
2019-03-03 23:21:58.336 27985-28793/com.example.ezminute I/System.out: "confidence": 0.579,
2019-03-03 23:21:58.336 27985-28793/com.example.ezminute I/System.out: "final": false,
2019-03-03 23:21:58.336 27985-28793/com.example.ezminute I/System.out: "from": 4.15,
2019-03-03 23:21:58.336 27985-28793/com.example.ezminute I/System.out: "speaker": 0,
2019-03-03 23:21:58.336 27985-28793/com.example.ezminute I/System.out: "to": 4.69
2019-03-03 23:21:58.336 27985-28793/com.example.ezminute I/System.out: }
2019-03-03 23:21:58.336 27985-28793/com.example.ezminute I/System.out: ]
2019-03-03 23:21:58.336 27985-28793/com.example.ezminute I/System.out: }
2019-03-03 23:21:58.340 4354-4354/? D/io_stats: !@ 179,0 r 137002 5177089 w 48914 1018780 d 7995 317604 f 12673 12674 iot 105310 98590 th 51200 0 0 pt 0 inp 0 0 3948.151
2019-03-03 23:21:58.345 27985-28793/com.example.ezminute I/System.out: {
2019-03-03 23:21:58.345 27985-28793/com.example.ezminute I/System.out: "speaker_labels": [
2019-03-03 23:21:58.345 27985-28793/com.example.ezminute I/System.out: {
2019-03-03 23:21:58.346 27985-28793/com.example.ezminute I/System.out: "confidence": 0.579,
2019-03-03 23:21:58.346 27985-28793/com.example.ezminute I/System.out: "final": true,
2019-03-03 23:21:58.346 27985-28793/com.example.ezminute I/System.out: "from": 4.15,
2019-03-03 23:21:58.346 27985-28793/com.example.ezminute I/System.out: "speaker": 0,
2019-03-03 23:21:58.346 27985-28793/com.example.ezminute I/System.out: "to": 4.69
2019-03-03 23:21:58.346 27985-28793/com.example.ezminute I/System.out: }
2019-03-03 23:21:58.346 27985-28793/com.example.ezminute I/System.out: ]
2019-03-03 23:21:58.346 27985-28793/com.example.ezminute I/System.out: }
这是演示中的结果
{
"speaker_labels": [
{
"from": 2.43,
"to": 2.93,
"speaker": 0,
"confidence": 0.605,
"final": false
},
{
"from": 3.3,
"to": 3.82,
"speaker": 1,
"confidence": 0.667,
"final": false
},
{
"from": 4.15,
"to": 4.69,
"speaker": 0,
"confidence": 0.579,
"final": false
}
]
}
{
"speaker_labels": [
{
"from": 4.15,
"to": 4.69,
"speaker": 0,
"confidence": 0.579,
"final": true
}
]
}
这是我的输出
Speaker0: hello hi hello
这是演示中的输出
Speaker 0:
Hello.
Speaker 1:
Hi.
Speaker 0:
Hello.
请帮助我请恳求你们所有人请请帮助我:( :( :(我不知道该怎么办请尽我所能但是还不够我需要所有帮助请帮助我:( :( :(
有了这么小的音轨，我假设onTranscription只被调用一次。
我认为你的目的是利用SpeakerLabelsDiarization.RecoTokens来确定谁说什么和何时，但你没有使用它。您所做的只是实例化一个实例
SpeakerLabelsDiarization.RecoTokens recoTokens = new SpeakerLabelsDiarization.RecoTokens();
并添加STT响应
recoTokens.add(speechResults);
但随后onTranscription结束，recoTokens超出范围并消失。使整个练习毫无意义。
你的输出来自
String speaks = "Speaker " + speechResults.getSpeakerLabels().get(0).getSpeaker().toString();
这直接来自您从STT服务获得的响应。 get(0)将获得并允许您打印数组中的第0个（第一个）元素。由于您没有迭代，这就是您要打印的全部内容。
如果你想看到所有扬声器标签，你需要一个for循环来通过响应中的所有扬声器标签。
您输出的成绩单是完整的成绩单，不会被演讲者标签或时间细分。
String text = speechResults.getResults().get(0).getAlternatives().get(0).getTranscript();
我想你想要使用你已经建立的recoTokens，在简短的检查中会使用
recoTokens.uttterances()