SpeechPlugin.m 20 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429
  1. #import "SpeechPlugin.h"
  2. #import <iflyMSC/iflyMSC.h>
  3. #import "M4aToPcmHelper.h"
  4. #import "Mp4ToPcmHelper.h"
  5. #import "Results/ISEResult.h"
  6. #import "Results/ISEResultXmlParser.h"
  7. #import "Results/ISEResultTools.h"
  8. #import "aiengine.h"
  9. #import <TAISDK/TAISDK.h>
  10. #import <TAISDK/TAIOralEvaluation.h>
  11. @interface SpeechPlugin () <IFlySpeechEvaluatorDelegate, ISEResultXmlParserDelegate,TAIOralEvaluationDelegate>
  12. @property (nonatomic, strong) IFlySpeechEvaluator *iFlySpeechEvaluator;
  13. @property (nonatomic, strong) NSNumber *index;
  14. @property struct aiengine * engine;
  15. @property (strong, nonatomic) TAIOralEvaluation *oralEvaluation;
  16. @end
  17. @implementation SpeechPlugin
  18. + (void)registerWithRegistrar:(NSObject<FlutterPluginRegistrar>*)registrar {
  19. _channel = [FlutterMethodChannel
  20. methodChannelWithName:@"speech_plugin"
  21. binaryMessenger:[registrar messenger]];
  22. SpeechPlugin* instance = [[SpeechPlugin alloc] init];
  23. [registrar addMethodCallDelegate:instance channel: _channel];
  24. }
  25. - (void)handleMethodCall:(FlutterMethodCall*)call result:(FlutterResult)result {
  26. if ([@"getPlatformVersion" isEqualToString:call.method]) {
  27. result([@"iOS " stringByAppendingString:[[UIDevice currentDevice] systemVersion]]);
  28. } else if ([@"initSpeechSdk" isEqualToString:call.method]) {
  29. [self iflyInit];
  30. self.oralEvaluation=[[TAIOralEvaluation alloc]init];
  31. self.oralEvaluation.delegate=self;
  32. } else if ([@"evaluatorByAudio" isEqualToString:call.method]) {
  33. NSNumber* index = call.arguments[@"index"];
  34. NSString* recordPath = call.arguments[@"recordPath"];
  35. NSString* text = call.arguments[@"en"];
  36. NSString* pathEvaluatorDecode = call.arguments[@"pathEvaluatorDecode"];
  37. NSString* videoId = call.arguments[@"videoId"];
  38. NSNumber* evaluatorType = call.arguments[@"evaluatorType"];
  39. NSNumber* sdkType = call.arguments[@"sdkType"];
  40. [self evaluateVoice:index andPath:recordPath andText:text andIsVideo:false pathEvaluatorDecode:pathEvaluatorDecode videoId:videoId evaluatorType:evaluatorType sdkType:sdkType];
  41. } else if ([@"evaluatorByMp4" isEqualToString:call.method]) {
  42. NSNumber* index = call.arguments[@"index"];
  43. NSString* recordPath = call.arguments[@"recordPath"];
  44. NSString* text = call.arguments[@"en"];
  45. NSString* pathEvaluatorDecode = call.arguments[@"pathEvaluatorDecode"];
  46. NSString* videoId = call.arguments[@"videoId"];
  47. NSNumber* evaluatorType = call.arguments[@"evaluatorType"];
  48. NSNumber* sdkType = call.arguments[@"sdkType"];
  49. [self evaluateVoice:index andPath:recordPath andText:text andIsVideo:true
  50. pathEvaluatorDecode:pathEvaluatorDecode videoId:videoId evaluatorType:evaluatorType sdkType:sdkType];
  51. } else {
  52. result(FlutterMethodNotImplemented);
  53. }
  54. }
  55. #pragma mark - Bridge Actions
  56. - (void)iflyInit {
  57. [IFlySpeechUtility createUtility:@"appid=5db7af6b"];
  58. self.iFlySpeechEvaluator = [IFlySpeechEvaluator sharedInstance];
  59. self.iFlySpeechEvaluator.delegate = self;
  60. [self configEvaluator];
  61. }
  62. - (void)chivosInit:(NSString *)en
  63. {
  64. char cfg[4096];
  65. char version[512] = {0};
  66. char record_id[64] = {0};
  67. char param[4096];
  68. NSString* user_id = @"ios_user";
  69. /*获取当前SDK版本号*/
  70. aiengine_opt(NULL, AIENGINE_OPT_GET_VERSION, version, sizeof(version));
  71. NSLog(@"version: %s\n",version);
  72. /*获取证书路径*/
  73. NSString * provision = [[NSBundle mainBundle] pathForResource:@"aiengine" ofType:@"provision"];
  74. /*引擎初始化传参设置*/
  75. NSMutableDictionary *jsonDic = [[NSMutableDictionary alloc] initWithCapacity:10];
  76. /*授权参数*/
  77. [jsonDic setValue:@"157775873600002d" forKey:@"appKey"];
  78. [jsonDic setValue:@"a6c766845f9a83974aed16f103e60621" forKey:@"secretKey"];
  79. [jsonDic setValue:provision forKey:@"provision"];
  80. /*日志传参*/
  81. NSMutableDictionary *logDic = [[NSMutableDictionary alloc] init];
  82. NSString *logFileName = @"log.txt";
  83. [logDic setValue:[NSNumber numberWithInt:0] forKey:@"enable"];
  84. [logDic setValue:logFileName forKey:@"output"];
  85. /*服务链接参数*/
  86. NSMutableDictionary *cloudDic = [[NSMutableDictionary alloc] init];
  87. NSString *serverPath = @"wss://cloud.chivox.com:443";
  88. [cloudDic setValue:[NSNumber numberWithInt:1] forKey:@"enable"];
  89. [cloudDic setValue:serverPath forKey:@"server"];
  90. [cloudDic setValue:[NSNumber numberWithInt:60] forKey:@"serverTimeout"];
  91. /*初始化传参设置*/
  92. [jsonDic setValue:logDic forKey:@"prof"];
  93. [jsonDic setValue:cloudDic forKey:@"cloud"];
  94. NSString *jsonString;
  95. jsonString = [self dicToString:jsonDic];
  96. /*cfg赋值:string转char*/
  97. strcpy(cfg, [jsonString UTF8String]);
  98. NSLog(@"cfg: %s\n",cfg);
  99. /*评分引擎初始化*/
  100. _engine = aiengine_new(cfg);
  101. NSLog(@"engine: %p\n", _engine);
  102. /*引擎启用param传参设置*/
  103. NSMutableDictionary *paramDic = [[NSMutableDictionary alloc] initWithCapacity:10];
  104. /*在线离线参数配置(cloud/native)*/
  105. [paramDic setValue:@"cloud" forKey:@"coreProvideType"];
  106. /*音量实时返回参数设置(0关/1开)*/
  107. [paramDic setValue:[NSNumber numberWithInt:0] forKey:@"soundIntensityEnable"];
  108. /*appUser传参*/
  109. NSMutableDictionary *appDic = [[NSMutableDictionary alloc] init];
  110. [appDic setValue:user_id forKey:@"userId"];
  111. /*audio音频数据传参*/
  112. NSMutableDictionary *audioDic = [[NSMutableDictionary alloc] init];
  113. [audioDic setValue:@"wav" forKey:@"audioType"];//音频编码格式
  114. [audioDic setValue:[NSNumber numberWithInt:16000] forKey:@"sampleRate"];//音频采样率
  115. [audioDic setValue:[NSNumber numberWithInt:1] forKey:@"channel"];//单声道设置
  116. [audioDic setValue:[NSNumber numberWithInt:2] forKey:@"sampleBytes"];//采样字节数(1-单字节-8位,2-双字节-16位)
  117. /*内核request传参*/
  118. NSMutableDictionary *requestDic = [[NSMutableDictionary alloc] init];
  119. //内核类型设置cn.word.raw/cn.sent.raw/cn.pred.raw
  120. [requestDic setValue:@"en.sent.score" forKey:@"coreType"];
  121. [requestDic setValue:en forKey:@"refText"];//评测文本
  122. [requestDic setValue:[NSNumber numberWithInt:100] forKey:@"rank"];//总分分制
  123. [requestDic setValue:[NSNumber numberWithInt:1] forKey:@"attachAudioUrl"];
  124. [requestDic setValue:[NSNumber numberWithInt:1] forKey:@"precision"];//段落评分精度设置(0.5/1)
  125. [paramDic setValue:appDic forKey:@"app"];
  126. [paramDic setValue:audioDic forKey:@"audio"];
  127. [paramDic setValue:requestDic forKey:@"request"];
  128. NSString *paramString;
  129. paramString = [self dicToString:paramDic];
  130. /*cfg赋值:string转char*/
  131. strcpy(param, [paramString UTF8String]);
  132. int rv = 0;
  133. rv = aiengine_start(_engine,param,record_id,(aiengine_callback)_aiengine_callback, (__bridge const void *)(self));
  134. if (rv) {
  135. NSLog(@"aiengine_start() failed: %d\n", rv);
  136. return;
  137. }
  138. }
  139. /*字典转为String*/
  140. - (NSString *)dicToString: (NSMutableDictionary *) jsonDictionary{
  141. NSError *error;
  142. NSData *jsonData = [NSJSONSerialization dataWithJSONObject:jsonDictionary options:0 error:&error];
  143. NSString *jsString;
  144. if (!jsonData) {
  145. NSLog(@"sorry you get an error:%@",error);
  146. } else {
  147. jsString = [[NSString alloc]initWithData:jsonData encoding:NSUTF8StringEncoding];
  148. }
  149. NSLog(@"引擎初始化传参#####:%@",jsString);
  150. return jsString;
  151. }
  152. - (void) configEvaluator {
  153. [self.iFlySpeechEvaluator setParameter:@"" forKey:[IFlySpeechConstant PARAMS]];
  154. [self.iFlySpeechEvaluator setParameter:@"read_sentence" forKey:[IFlySpeechConstant ISE_CATEGORY]];
  155. [self.iFlySpeechEvaluator setParameter:@"en_us" forKey:[IFlySpeechConstant LANGUAGE]];
  156. [self.iFlySpeechEvaluator setParameter:@"5000" forKey:[IFlySpeechConstant VAD_BOS]];
  157. [self.iFlySpeechEvaluator setParameter:@"1800" forKey:[IFlySpeechConstant VAD_EOS]];
  158. [self.iFlySpeechEvaluator setParameter:@"-1" forKey:[IFlySpeechConstant SPEECH_TIMEOUT]];
  159. [self.iFlySpeechEvaluator setParameter:@"complete" forKey:[IFlySpeechConstant ISE_RESULT_LEVEL]];
  160. [self.iFlySpeechEvaluator setParameter:@"16000" forKey:[IFlySpeechConstant SAMPLE_RATE]];
  161. [self.iFlySpeechEvaluator setParameter:@"xml" forKey:[IFlySpeechConstant ISE_RESULT_TYPE]];
  162. [self.iFlySpeechEvaluator setParameter:@"0" forKey:@"plev"];
  163. [self.iFlySpeechEvaluator setParameter:@"-1" forKey:@"audio_source"];
  164. }
  165. - (void) evaluateVoice: (NSNumber*)index andPath:(NSString*)path andText:(NSString*)text andIsVideo:(BOOL) isVideo pathEvaluatorDecode:(NSString*)pathEvaluatorDecode videoId:(NSString*)videoId evaluatorType:(NSNumber*)evaluatorType sdkType:(NSNumber*)sdkType
  166. {
  167. self.index = index;
  168. if (sdkType.intValue == 0) {
  169. //调用讯飞解析
  170. if(isVideo) {
  171. [Mp4ToPcmHelper Mp4ToPcmWithUrl:[[NSURL alloc] initFileURLWithPath:path] completion:^(NSData *data) {
  172. if(data == nil) {
  173. [_channel invokeMethod:@"evaluatorResult" arguments: [NSDictionary dictionaryWithObjectsAndKeys: self.index, @"index", [NSNull null], @"score", nil]];
  174. return;
  175. }
  176. NSStringEncoding encoding = CFStringConvertEncodingToNSStringEncoding(kCFStringEncodingGB_18030_2000);
  177. NSMutableData *buffer= [NSMutableData dataWithData:[text dataUsingEncoding:encoding]];
  178. [self.iFlySpeechEvaluator startListening:buffer params:nil];
  179. [self.iFlySpeechEvaluator writeAudio:data];
  180. [self.iFlySpeechEvaluator stopListening];
  181. }];
  182. } else {
  183. NSData *voiceData = [M4aToPcmHelper M4aToPcmWithUrl:[[NSURL alloc] initFileURLWithPath:path]];
  184. if(voiceData == nil) {
  185. [_channel invokeMethod:@"evaluatorResult" arguments: [NSDictionary dictionaryWithObjectsAndKeys: self.index, @"index", [NSNull null], @"score", nil]];
  186. return;
  187. }
  188. NSStringEncoding encoding = CFStringConvertEncodingToNSStringEncoding(kCFStringEncodingGB_18030_2000);
  189. NSMutableData *buffer= [NSMutableData dataWithData:[text dataUsingEncoding:encoding]];
  190. [self.iFlySpeechEvaluator startListening:buffer params:nil];
  191. [self.iFlySpeechEvaluator writeAudio:voiceData];
  192. [self.iFlySpeechEvaluator stopListening];
  193. }
  194. }else if(sdkType.intValue == 1){
  195. //调用池声解析
  196. [self chivosInit:text];
  197. if(isVideo){
  198. }else {
  199. int bytes = 0;
  200. char buf[1024]={0};
  201. FILE *file = NULL;
  202. const char * audiopath = [path UTF8String];
  203. file = fopen(audiopath, "rb");//待评分音频地址
  204. if(!file)
  205. {
  206. printf("read file error!\n");
  207. return;
  208. }
  209. // NSData *voiceData = [M4aToPcmHelper M4aToPcmWithUrl:[[NSURL alloc] initFileURLWithPath:path]];
  210. // NSUInteger len = [voiceData length];
  211. // Byte *bytedata = (Byte*)malloc(len);
  212. // memcpy(bytedata, [voiceData bytes], len);
  213. //
  214. // aiengine_feed(_engine, bytedata, len);
  215. // fseek(file, 44, SEEK_SET); //wav音频文件需要跳过头文件信息,其它音频格式不需要
  216. printf("startFeed\n");
  217. while ((bytes = (int)fread(buf, 1, 1024, file)))
  218. {
  219. aiengine_feed(_engine, buf, bytes);
  220. }
  221. aiengine_stop(_engine);
  222. }
  223. }else {
  224. //调用腾讯解析
  225. TAIOralEvaluationParam *param =[[TAIOralEvaluationParam alloc]init];
  226. param.sessionId = [[NSUUID UUID]UUIDString];
  227. param.appId =@"1301049120";
  228. param.workMode = TAIOralEvaluationWorkMode_Once;
  229. param.evalMode = TAIOralEvaluationEvalMode_Sentence;
  230. param.storageMode = TAIOralEvaluationStorageMode_Disable;
  231. param.serverType = TAIOralEvaluationServerType_English;
  232. param.scoreCoeff = 1.0;
  233. param.fileType = TAIOralEvaluationFileType_Wav;
  234. param.refText = text;
  235. param.secretId = @"AKIDUf0EEzc8NMpPmu4Po1zhXBZicKp5G7xZ";
  236. param.secretKey = @"WnsjKaqtgDVbV9cMtABwVptarwpWyBAt";
  237. TAIOralEvaluationData *data = [[TAIOralEvaluationData alloc] init];
  238. data.seqId = 1;
  239. data.bEnd = YES;
  240. data.audio = [NSData dataWithContentsOfFile:path];
  241. [self.oralEvaluation oralEvaluation:param data:data callback:^(TAIError *error) {
  242. //接口调用结果返回
  243. if(error){
  244. NSLog(@"调用腾讯失败:code:%ld , msg:%@",(long)error.code ,error.desc);
  245. }else{
  246. NSLog(@"调用腾讯成功");
  247. }
  248. }];
  249. }
  250. }
  251. #pragma mark - iFly delegate
  252. // 评测结果回调
  253. - (void)onResults:(NSData *)results isLast:(BOOL)isLast {
  254. if (isLast) {
  255. if(results) {
  256. const char* chResult = [results bytes];
  257. NSString* strResults = nil;
  258. NSStringEncoding encoding = CFStringConvertEncodingToNSStringEncoding(kCFStringEncodingGB_18030_2000);
  259. strResults = [[NSString alloc] initWithBytes:chResult length:[results length] encoding:encoding];
  260. if(strResults != nil) {
  261. ISEResultXmlParser *parser = [ISEResultXmlParser alloc];
  262. [parser setDelegate:self];
  263. [parser parserXml:strResults];
  264. } else {
  265. [_channel invokeMethod:@"evaluatorResult" arguments: [NSDictionary dictionaryWithObjectsAndKeys: self.index, @"index", [NSNull null], @"score", nil]];
  266. }
  267. } else {
  268. [_channel invokeMethod:@"evaluatorResult" arguments: [NSDictionary dictionaryWithObjectsAndKeys: self.index, @"index", [NSNull null], @"score", nil]];
  269. }
  270. }
  271. }
  272. - (void)onCompleted:(IFlySpeechError *)errorCode {
  273. if (errorCode.errorCode != 0) {
  274. [_channel invokeMethod:@"evaluatorResult" arguments: [NSDictionary dictionaryWithObjectsAndKeys: self.index, @"index", [NSNull null], @"score", nil]];
  275. }
  276. }
  277. - (void)onCancel {}
  278. - (void)onBeginOfSpeech {}
  279. - (void)onEndOfSpeech {}
  280. - (void)onVolumeChanged:(int)volume buffer:(NSData *)buffer {}
  281. #pragma mark - ISEResultXmlParser delegate
  282. -(void)onISEResultXmlParser:(NSXMLParser *)parser Error:(NSError*)error {
  283. if (error.code != 0) {
  284. [_channel invokeMethod:@"evaluatorResult" arguments: [NSDictionary dictionaryWithObjectsAndKeys: self.index, @"index", [NSNull null], @"score", nil]];
  285. }
  286. }
  287. -(void)onISEResultXmlParserResult:(ISEResult*)result {
  288. if (result.is_rejected) {
  289. [_channel invokeMethod:@"evaluatorResult" arguments: [NSDictionary dictionaryWithObjectsAndKeys: self.index, @"index", [NSNull null], @"score", nil]];
  290. } else {
  291. NSMutableDictionary *dic = [NSMutableDictionary dictionary];
  292. [dic setValue:self.index forKey:@"index"];
  293. [dic setValue:@(result.total_score) forKey:@"score"];
  294. [dic setValue:@(result.accuracy_score) forKey:@"accuracy_score"];
  295. [dic setValue:@(result.fluency_score) forKey:@"fluency_score"];
  296. [dic setValue:@(result.integrity_score) forKey:@"integrity_score"];
  297. NSString* words = [ISEResultTools formatDetailsForLanguageEN: result.sentences];
  298. [dic setValue:words forKey:@"words"];
  299. [_channel invokeMethod:@"evaluatorResult" arguments: dic];
  300. }
  301. }
  302. #pragma mark - chisheng回掉
  303. /*评分引擎回调方法*/
  304. int _aiengine_callback(const void *usrdata, const char *id, int type, const void *message, int size)
  305. {
  306. if (type == AIENGINE_MESSAGE_TYPE_JSON) {
  307. [(__bridge SpeechPlugin *)usrdata performSelectorOnMainThread:@selector(onChiResult:) withObject:[[NSString alloc] initWithUTF8String:(char *)message] waitUntilDone:NO];
  308. }
  309. return 0;
  310. }
  311. //返回的内容
  312. -(void) onChiResult:(NSString *) result{
  313. if(result == nil){
  314. [_channel invokeMethod:@"evaluatorResult" arguments: [NSDictionary dictionaryWithObjectsAndKeys: self.index, @"index", [NSNull null], @"score", nil]];
  315. return;
  316. }
  317. NSLog(@"返回的结果:%@",result);
  318. NSData *jsonData = [result dataUsingEncoding:NSUTF8StringEncoding];
  319. NSError *err = nil;
  320. NSMutableDictionary *dic = [NSJSONSerialization JSONObjectWithData:jsonData options:NSJSONReadingMutableContainers error:&err];
  321. if(err) {
  322. NSLog(@"json解析失败:%@",err);
  323. [_channel invokeMethod:@"evaluatorResult" arguments: [NSDictionary dictionaryWithObjectsAndKeys: self.index, @"index", [NSNull null], @"score", nil]];
  324. return ;
  325. }
  326. NSNumber *score = dic[@"result"][@"overall"];
  327. NSNumber *integrity = dic[@"result"][@"integrity"];
  328. NSNumber *accuracy = dic[@"result"][@"accuracy"];
  329. NSNumber *fluency = dic[@"result"][@"fluency"][@"overall"];
  330. NSArray<NSDictionary *>* worksList = dic[@"result"][@"details"];
  331. NSMutableDictionary *dict = [NSMutableDictionary dictionary];
  332. [dict setValue:self.index forKey:@"index"];
  333. [dict setValue:@(score.doubleValue/20) forKey:@"score"];
  334. [dict setValue:@(accuracy.doubleValue/20) forKey:@"accuracy_score"];
  335. [dict setValue:@(fluency.doubleValue/20) forKey:@"fluency_score"];
  336. [dict setValue:@(integrity.doubleValue/20) forKey:@"integrity_score"];
  337. [dict setValue:worksList forKey:@"words"];
  338. [_channel invokeMethod:@"evaluatorResult" arguments: dict];
  339. if (_engine) {
  340. aiengine_delete(_engine);
  341. _engine = NULL;
  342. }
  343. }
  344. #pragma mark - 腾讯会掉
  345. - (void)oralEvaluation:(TAIOralEvaluation *)oralEvaluation onEvaluateData:(TAIOralEvaluationData *)data result:(TAIOralEvaluationRet *)result error:(TAIError *)error {
  346. //数据和结果回调(只有data.bEnd为YES,result有效)
  347. if (data.bEnd) {
  348. float pronFluency = result.pronFluency *5;
  349. float pronAccuracy = result.pronAccuracy;
  350. if(pronAccuracy == -1){
  351. pronAccuracy = 0;
  352. }else {
  353. pronAccuracy = pronAccuracy/20;
  354. }
  355. float pronCompletion = result.pronCompletion * 5;
  356. float totalScore = (pronFluency + pronAccuracy + pronCompletion)/ 3;
  357. NSMutableArray* workList =[NSMutableArray array];
  358. for (TAIOralEvaluationWord* work in result.words) {
  359. NSMutableDictionary *dic =[NSMutableDictionary dictionary];
  360. float score = work.pronAccuracy;
  361. if(score == -1.0){
  362. score = 0;
  363. }else{
  364. score = score/20;
  365. }
  366. dic[@"score"]= [NSNumber numberWithFloat:score];
  367. dic[@"content"] = work.word;
  368. [workList addObject:dic];
  369. }
  370. NSMutableDictionary *dict = [NSMutableDictionary dictionary];
  371. [dict setValue:self.index forKey:@"index"];
  372. [dict setValue:@(totalScore) forKey:@"score"];
  373. [dict setValue:@(pronAccuracy) forKey:@"accuracy_score"];
  374. [dict setValue:@(pronFluency) forKey:@"fluency_score"];
  375. [dict setValue:@(pronCompletion) forKey:@"integrity_score"];
  376. [dict setValue:workList forKey:@"words"];
  377. [_channel invokeMethod:@"evaluatorResult" arguments: dict];
  378. }else{
  379. [_channel invokeMethod:@"evaluatorResult" arguments: [NSDictionary dictionaryWithObjectsAndKeys: self.index, @"index", [NSNull null], @"score", nil]];
  380. }
  381. }
  382. @end