ASRView.m 15 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382
  1. //
  2. // ASRView.m
  3. // YBVideo
  4. //
  5. // Created by YunBao on 2018/7/31.
  6. // Copyright © 2018年 cat. All rights reserved.
  7. //
  8. #import "ASRView.h"
  9. #import "BDSASRDefines.h"
  10. #import "BDSASRParameters.h"
  11. #import "BDSWakeupDefines.h"
  12. #import "BDSWakeupParameters.h"
  13. #import "BDSEventManager.h"
  14. #import "BDRecognizerViewController.h"
  15. #import "BDVRSettings.h"
  16. #import "fcntl.h"
  17. #import "AudioInputStream.h"
  18. @interface ASRView()<UITextViewDelegate,BDSClientASRDelegate>
  19. {
  20. NSString *_endRsStr; //识别字段最终结果
  21. }
  22. /** 输入框、返回键、清除按钮、发送按钮 */
  23. @property(nonatomic,strong)UIView *botMixView;
  24. //@property(nonatomic,strong)MyTextView *textView; //输入框 .h
  25. @property(nonatomic,strong)UILabel *explainL; //按住说话
  26. @property(nonatomic,strong)UIButton *backBtn; //返回按钮
  27. @property(nonatomic,strong)UIButton *clearBtn; //清除按钮
  28. @property(nonatomic,strong)UIButton *sendBtn; //发送按钮
  29. @property(nonatomic,strong)UIButton *voiceBtn; //录音按钮
  30. @property (strong, nonatomic) BDSEventManager *asrEventManager;
  31. @property(nonatomic, assign) BOOL continueToVR;
  32. @property(nonatomic, strong) NSFileHandle *fileHandler;
  33. @property(nonatomic, strong) BDRecognizerViewController *recognizerViewController;
  34. @property(nonatomic, assign) TBDVoiceRecognitionOfflineEngineType curOfflineEngineType;
  35. @property(nonatomic, strong) NSTimer *longPressTimer;
  36. @property(nonatomic, assign) BOOL longPressFlag;
  37. @property(nonatomic, assign) BOOL touchUpFlag;
  38. @property(nonatomic, assign) BOOL longSpeechFlag;
  39. @end
  40. @implementation ASRView
  41. - (instancetype)initWithFrame:(CGRect)frame callBack:(ASRBlock)asrBack{
  42. self = [super initWithFrame:frame];
  43. if (self) {
  44. self.asrEvent = asrBack;
  45. _endRsStr = [NSString string];
  46. [self addSubview:self.botMixView];
  47. self.asrEventManager = [BDSEventManager createEventManagerWithName:BDS_ASR_NAME];
  48. NSLog(@"Current SDK version: %@", [self.asrEventManager libver]);
  49. self.continueToVR = NO;
  50. [[BDVRSettings getInstance] configBDVRClient];
  51. [self configVoiceRecognitionClient];
  52. }
  53. return self;
  54. }
  55. -(void)dealloc {
  56. [[NSNotificationCenter defaultCenter]removeObserver:self];
  57. }
  58. #pragma mark - 点击事件
  59. //按下
  60. -(void)clickVoiceBtnDown:(UIButton *)btn {
  61. [btn.layer addAnimation:[PublicObj touchDownAnimation] forKey:nil];
  62. self.touchUpFlag = NO;
  63. self.longPressFlag = NO;
  64. self.longPressTimer = [NSTimer timerWithTimeInterval:0.5
  65. target:self
  66. selector:@selector(longPressTimerTriggered) userInfo:nil repeats:NO];
  67. [[NSRunLoop currentRunLoop] addTimer:self.longPressTimer forMode:NSRunLoopCommonModes];
  68. [self.asrEventManager setParameter:@(NO) forKey:BDS_ASR_ENABLE_LONG_SPEECH];
  69. [self.asrEventManager setParameter:@(NO) forKey:BDS_ASR_NEED_CACHE_AUDIO];
  70. [self.asrEventManager setParameter:@"" forKey:BDS_ASR_OFFLINE_ENGINE_TRIGGERED_WAKEUP_WORD];
  71. [self voiceRecogButtonHelper];
  72. }
  73. - (void)voiceRecogButtonHelper {
  74. //[self configFileHandler];
  75. [self.asrEventManager setDelegate:self];
  76. [self.asrEventManager setParameter:nil forKey:BDS_ASR_AUDIO_FILE_PATH];
  77. [self.asrEventManager setParameter:nil forKey:BDS_ASR_AUDIO_INPUT_STREAM];
  78. [self.asrEventManager sendCommand:BDS_ASR_CMD_START];
  79. }
  80. - (void)longPressTimerTriggered {
  81. if (!self.touchUpFlag) {
  82. self.longPressFlag = YES;
  83. [self.asrEventManager setParameter:@(YES) forKey:BDS_ASR_VAD_ENABLE_LONG_PRESS];
  84. }
  85. [self.longPressTimer invalidate];
  86. }
  87. //抬起
  88. -(void)clickVoiceBtnUp:(UIButton *)btn {
  89. [btn.layer removeAllAnimations];
  90. self.touchUpFlag = YES;
  91. if (self.longPressFlag) {
  92. [self.asrEventManager sendCommand:BDS_ASR_CMD_STOP];
  93. }
  94. }
  95. //返回
  96. -(void)clickBackBtn {
  97. if (self.asrEvent) {
  98. self.asrEvent(@"返回", @"");
  99. }
  100. }
  101. //清除
  102. -(void)clickClearBtn {
  103. _endRsStr = @"";
  104. _textView.text = @"";
  105. [self noContent];
  106. }
  107. //发送
  108. -(void)clickSendBtn {
  109. if (self.asrEvent) {
  110. self.asrEvent(@"发送", _textView.text);
  111. [self clickClearBtn];
  112. }
  113. }
  114. -(void)haveContent {
  115. _backBtn.hidden = YES;
  116. _clearBtn.hidden = NO;
  117. _sendBtn.hidden = NO;
  118. }
  119. -(void)noContent {
  120. _backBtn.hidden = NO;
  121. _clearBtn.hidden = YES;
  122. _sendBtn.hidden = YES;
  123. }
  124. #pragma mark - textView 代理
  125. #pragma mark - Private: Configuration
  126. - (void)configVoiceRecognitionClient {
  127. //设置DEBUG_LOG的级别
  128. [self.asrEventManager setParameter:@(EVRDebugLogLevelTrace) forKey:BDS_ASR_DEBUG_LOG_LEVEL];
  129. //配置API_KEY 和 SECRET_KEY 和 APP_ID
  130. [self.asrEventManager setParameter:@[ASR_API_KEY, ASR_SECRET_KEY] forKey:BDS_ASR_API_SECRET_KEYS];
  131. [self.asrEventManager setParameter:ASR_APP_ID forKey:BDS_ASR_OFFLINE_APP_CODE];
  132. //配置端点检测(二选一)
  133. // [self configModelVAD];
  134. [self configDNNMFE];
  135. //[self.asrEventManager setParameter:@"15361" forKey:BDS_ASR_PRODUCT_ID];
  136. // ---- 语义与标点 -----
  137. //[self enableNLU];
  138. [self enablePunctuation];
  139. // ------------------------
  140. }
  141. - (void) enableNLU {
  142. // ---- 开启语义理解 -----
  143. [self.asrEventManager setParameter:@(YES) forKey:BDS_ASR_ENABLE_NLU];
  144. [self.asrEventManager setParameter:@"1536" forKey:BDS_ASR_PRODUCT_ID];
  145. }
  146. - (void) enablePunctuation {
  147. // ---- 开启标点输出 -----
  148. [self.asrEventManager setParameter:@(NO) forKey:BDS_ASR_DISABLE_PUNCTUATION];
  149. // 普通话标点
  150. [self.asrEventManager setParameter:@"1537" forKey:BDS_ASR_PRODUCT_ID];
  151. // 英文标点
  152. //[self.asrEventManager setParameter:@"1737" forKey:BDS_ASR_PRODUCT_ID];
  153. }
  154. - (void)configModelVAD {
  155. NSString *modelVAD_filepath = [[NSBundle mainBundle] pathForResource:@"bds_easr_basic_model" ofType:@"dat"];
  156. [self.asrEventManager setParameter:modelVAD_filepath forKey:BDS_ASR_MODEL_VAD_DAT_FILE];
  157. [self.asrEventManager setParameter:@(YES) forKey:BDS_ASR_ENABLE_MODEL_VAD];
  158. }
  159. - (void)configDNNMFE {
  160. NSString *mfe_dnn_filepath = [[NSBundle mainBundle] pathForResource:@"bds_easr_mfe_dnn" ofType:@"dat"];
  161. [self.asrEventManager setParameter:mfe_dnn_filepath forKey:BDS_ASR_MFE_DNN_DAT_FILE];
  162. NSString *cmvn_dnn_filepath = [[NSBundle mainBundle] pathForResource:@"bds_easr_mfe_cmvn" ofType:@"dat"];
  163. [self.asrEventManager setParameter:cmvn_dnn_filepath forKey:BDS_ASR_MFE_CMVN_DAT_FILE];
  164. [self.asrEventManager setParameter:@(NO) forKey:BDS_ASR_ENABLE_MODEL_VAD];
  165. // MFE支持自定义静音时长
  166. [self.asrEventManager setParameter:@(5000.f) forKey:BDS_ASR_MFE_MAX_SPEECH_PAUSE];
  167. [self.asrEventManager setParameter:@(5000.f) forKey:BDS_ASR_MFE_MAX_WAIT_DURATION];
  168. }
  169. #pragma mark - MVoiceRecognitionClientDelegate
  170. - (void)VoiceRecognitionClientWorkStatus:(int)workStatus obj:(id)aObj {
  171. switch (workStatus) {
  172. case EVoiceRecognitionClientWorkStatusNewRecordData: {
  173. [self.fileHandler writeData:(NSData *)aObj];
  174. break;
  175. }case EVoiceRecognitionClientWorkStatusStartWorkIng: {
  176. _textView.placeholder = YZMsg(@"初始化中...");
  177. NSDictionary *logDic = [self parseLogToDic:aObj];
  178. [self printLogTextView:[NSString stringWithFormat:@"CALLBACK: start vr, log: %@\n", logDic]];
  179. // [self onStartWorking];
  180. break;
  181. }case EVoiceRecognitionClientWorkStatusStart: {
  182. _textView.placeholder = YZMsg(@"长按识别...");
  183. [self printLogTextView:@"CALLBACK: detect voice start point.\n"];
  184. break;
  185. }case EVoiceRecognitionClientWorkStatusEnd: {
  186. [self printLogTextView:@"CALLBACK: detect voice end point.\n"];
  187. break;
  188. }case EVoiceRecognitionClientWorkStatusFlushData: {
  189. [self printLogTextView:[NSString stringWithFormat:@"CALLBACK: partial result - %@.\n\n", [self getDescriptionObj:aObj]]];
  190. if (aObj) {
  191. _textView.text = [self getDescriptionObj:aObj];
  192. [self haveContent];
  193. }
  194. break;
  195. }case EVoiceRecognitionClientWorkStatusFinish: {
  196. [self printLogTextView:[NSString stringWithFormat:@"CALLBACK: final result - %@.\n\n", [self getDescriptionObj:aObj]]];
  197. // if (aObj) {
  198. // _textView.text = [_textView.text stringByAppendingString:[self getDescriptionObj:aObj]];
  199. // }
  200. if (!self.longSpeechFlag) {
  201. // [self onEnd];
  202. }
  203. break;
  204. }case EVoiceRecognitionClientWorkStatusMeterLevel: {
  205. break;
  206. }case EVoiceRecognitionClientWorkStatusCancel: {
  207. [self printLogTextView:@"CALLBACK: user press cancel.\n"];
  208. // [self onEnd];
  209. break;
  210. }case EVoiceRecognitionClientWorkStatusError: {
  211. [self printLogTextView:[NSString stringWithFormat:@"CALLBACK: encount error - %@.\n", (NSError *)aObj]];
  212. // [self onEnd];
  213. break;
  214. }case EVoiceRecognitionClientWorkStatusLoaded: {
  215. [self printLogTextView:@"CALLBACK: offline engine loaded.\n"];
  216. break;
  217. }case EVoiceRecognitionClientWorkStatusUnLoaded: {
  218. [self printLogTextView:@"CALLBACK: offline engine unLoaded.\n"];
  219. break;
  220. }case EVoiceRecognitionClientWorkStatusChunkThirdData: {
  221. [self printLogTextView:[NSString stringWithFormat:@"CALLBACK: Chunk 3-party data length: %lu\n", (unsigned long)[(NSData *)aObj length]]];
  222. break;
  223. }case EVoiceRecognitionClientWorkStatusChunkNlu: {
  224. NSString *nlu = [[NSString alloc] initWithData:(NSData *)aObj encoding:NSUTF8StringEncoding];
  225. [self printLogTextView:[NSString stringWithFormat:@"CALLBACK: Chunk NLU data: %@\n", nlu]];
  226. NSLog(@"%@", nlu);
  227. break;
  228. }case EVoiceRecognitionClientWorkStatusChunkEnd: {
  229. [self printLogTextView:[NSString stringWithFormat:@"CALLBACK: Chunk end, sn: %@.\n", aObj]];
  230. if (!self.longSpeechFlag) {
  231. // [self onEnd];
  232. }
  233. break;
  234. }case EVoiceRecognitionClientWorkStatusFeedback: {
  235. NSDictionary *logDic = [self parseLogToDic:aObj];
  236. [self printLogTextView:[NSString stringWithFormat:@"CALLBACK Feedback: %@\n", logDic]];
  237. break;
  238. }case EVoiceRecognitionClientWorkStatusRecorderEnd: {
  239. [self printLogTextView:@"CALLBACK: recorder closed.\n"];
  240. break;
  241. }case EVoiceRecognitionClientWorkStatusLongSpeechEnd: {
  242. [self printLogTextView:@"CALLBACK: Long Speech end.\n"];
  243. // [self onEnd];
  244. break;
  245. }default:
  246. break;
  247. }
  248. }
  249. - (NSDictionary *)parseLogToDic:(NSString *)logString {
  250. NSArray *tmp = NULL;
  251. NSMutableDictionary *logDic = [[NSMutableDictionary alloc] initWithCapacity:3];
  252. NSArray *items = [logString componentsSeparatedByString:@"&"];
  253. for (NSString *item in items) {
  254. tmp = [item componentsSeparatedByString:@"="];
  255. if (tmp.count == 2) {
  256. [logDic setObject:tmp.lastObject forKey:tmp.firstObject];
  257. }
  258. }
  259. return logDic;
  260. }
  261. - (void)printLogTextView:(NSString *)logString {
  262. NSLog(@"asr-log:%@",logString);
  263. }
  264. - (NSString *)getDescriptionObj:(id)obj {
  265. if (obj) {
  266. NSString *words = [NSString stringWithFormat:@"%@",[[obj valueForKey:@"results_recognition"] firstObject]];
  267. return words;
  268. }
  269. return YZMsg(@"解析错误");
  270. }
  271. - (void)textViewDidChange:(UITextView *)textView {
  272. if (textView.text.length>0) {
  273. [self haveContent];
  274. }else {
  275. [self noContent];
  276. }
  277. }
  278. #pragma mark - set/get
  279. - (UIView *)botMixView {
  280. if (!_botMixView) {
  281. _botMixView = [[UIView alloc]initWithFrame:CGRectMake(0, self.height-ASRHeight-ShowDiff, self.width, ASRHeight+ShowDiff)];
  282. _botMixView.backgroundColor =RGB(235, 235, 235);
  283. //输入框 15+100+20+15+80
  284. _textView = [[MyTextView alloc] initWithFrame:CGRectMake(10, 15, self.width-30, 110)];
  285. _textView.backgroundColor = [UIColor clearColor];
  286. _textView.delegate = self;
  287. _textView.font = SYS_Font(16);
  288. _textView.textColor = RGB_COLOR(@"#ffffff", 1);
  289. [_botMixView addSubview:_textView];
  290. _explainL = [[UILabel alloc]initWithFrame:CGRectMake(0, _textView.bottom, self.width, 20)];
  291. _explainL.text = YZMsg(@"按住说话");
  292. _explainL.textColor =UIColor.darkGrayColor;// RGB_COLOR(@"#8c8c8c", 1);
  293. _explainL.textAlignment = NSTextAlignmentCenter;
  294. [_botMixView addSubview:_explainL];
  295. _voiceBtn = [UIButton buttonWithType:UIButtonTypeCustom];
  296. _voiceBtn.frame = CGRectMake(self.width/2-40, _explainL.bottom+15, 80, 80);
  297. [_voiceBtn setImage:[UIImage imageNamed:@"asr_record"] forState:0];
  298. [_voiceBtn addTarget:self action:@selector(clickVoiceBtnDown:) forControlEvents:UIControlEventTouchDown];
  299. [_voiceBtn addTarget:self action:@selector(clickVoiceBtnUp:) forControlEvents:UIControlEventTouchUpInside|UIControlEventTouchUpOutside];
  300. [_botMixView addSubview:_voiceBtn];
  301. //返回按钮
  302. _backBtn = [UIButton buttonWithType:UIButtonTypeCustom];
  303. [_backBtn setImage:[UIImage imageNamed:@"asr_arrow"] forState:0];
  304. [_backBtn addTarget:self action:@selector(clickBackBtn) forControlEvents:UIControlEventTouchUpInside];
  305. [_botMixView addSubview:_backBtn];
  306. [_backBtn mas_makeConstraints:^(MASConstraintMaker *make) {
  307. make.width.mas_equalTo(48);
  308. make.height.mas_equalTo(24);
  309. make.left.mas_equalTo(_botMixView.mas_left).offset(30);
  310. make.centerY.mas_equalTo(_voiceBtn.mas_centerY);
  311. }];
  312. //取消
  313. _clearBtn = [UIButton buttonWithType:UIButtonTypeCustom];
  314. [_clearBtn setTitle:YZMsg(@"清除") forState:0];
  315. _clearBtn.titleLabel.font = SYS_Font(15);
  316. [_clearBtn setTitleColor:RGB_COLOR(@"#8c8c8c", 1) forState:0];
  317. [_clearBtn addTarget:self action:@selector(clickClearBtn) forControlEvents:UIControlEventTouchUpInside];
  318. [_botMixView addSubview:_clearBtn];
  319. _clearBtn.hidden = YES;
  320. [_clearBtn mas_makeConstraints:^(MASConstraintMaker *make) {
  321. make.width.height.mas_equalTo(60);
  322. make.left.mas_equalTo(_botMixView.mas_left).offset(30);
  323. make.centerY.mas_equalTo(_voiceBtn.mas_centerY);
  324. }];
  325. //发送
  326. _sendBtn = [UIButton buttonWithType:UIButtonTypeCustom];
  327. [_sendBtn setTitle:YZMsg(@"发送") forState:0];
  328. _sendBtn.titleLabel.font = SYS_Font(15);
  329. [_sendBtn setTitleColor:Pink_Cor forState:0];
  330. [_sendBtn addTarget:self action:@selector(clickSendBtn) forControlEvents:UIControlEventTouchUpInside];
  331. [_botMixView addSubview:_sendBtn];
  332. _sendBtn.hidden = YES;
  333. [_sendBtn mas_makeConstraints:^(MASConstraintMaker *make) {
  334. make.width.height.mas_equalTo(60);
  335. make.right.mas_equalTo(_botMixView.mas_right).offset(-30);
  336. make.centerY.mas_equalTo(_voiceBtn.mas_centerY);
  337. }];
  338. }
  339. return _botMixView;
  340. }
  341. @end