首页天道酬勤语音识别开源代码,esp8266语音识别

语音识别开源代码,esp8266语音识别

张世龙 05-12 05:28 8次浏览

espnet :端到端处理工具工具包文档网站: https://espnet.github.io/espnet/installation.html

github地址: https://github.com/espnet/espnet

paper:https://arxiv.org/pdf/1804.00015.pdf

整体代码结构espnet/# pythonmodulesutils/# utilityscriptsofespnettest/# unittest test _ utils/# unittestforexecutablescriptsegs thecompleterecipeforeachcorporaan4/# an4istinycorpusandcanbeobtainedfreeely, soitmightbesuitablefortutorialasr1/# ASR recipe-run.sh # executable script-cmd.sh # toselectthebackendforjobscheduler setupscriptforenvironmentvariables-conf/# containingconfigurationfiles-steps/# theutitied theutilsscriptsfromkaldi-utils theutilsscriptsfromkaldits1/# TTS recipe .1. espnetpython代码主要包含以下部分: 语音增强语音识别语言模型机器翻译(machine translation )语音翻译) speech translation

utils/路径下

数据格式处理add JSON.py : addmultiplejsonvaluestoaninputoroutputvaluechange _ YAML.py : changespecifiedattributesofayamange datributesofayamlfileconge 360 concatenatejsonfilesget _ YAML.py : getaspecifiedattributefromayamlfilejson 2sc TM.py 3: con gnizedjsontotextjson2trn _ mt.py : convertjsontomachinetranslationtranscriptionjson2trn.py : convertajsontoatranager wer dict.py : convertajsontoatranscriptionfilewithatokendictionarymergejson.py : merrgejson CP 2js on.py 3: giveneachfileplepson le : type.typecanbeomittedandthedefaultis“str”. mix-monter mixing wav.scpfilesintoamulti-channel wav.scpusingsox t.txtfiletojsonsplitjj fileforparallelprocessingtext2token.py : convertrawtexttotokenizedtextttext2vocabulary.py 3: crer nce.py : trimsliencewithsimplepowerthresholdingandmakesegmentsfile.(切割silence的框架) tr N2 CTM.py :转换传输工具ckpoints.py : averagemodelsfromsnapshot特征处理apply-cmvn.py : apply mean-variancenormalization-cmvn-stats.py : lizationstatisticsifwspecifierprovided 3360 per-utterancebydefault ifwxfilename :全局计算- f bank-feats.py : comppation ft-feats.py : computestftft k _ to _ wav.py : convertfbanktowavusingggriffin-lim algorithm (一种从频谱中获取语音时域信号的方法) cessingdump-PCM.py : dumppcmfilesfromawavscpfileeval-source-serce-seval ateeenhancedspeech.e.g./arg parse2rst.pyrefref.scpenh enh.scpoutdiroutputdiror ./doc/arg parse2rst.pyrefref.scpre F2.scpenh enh.sc Penh2. scpoutdiroutputdirfeats 2n py.py : convet kaldi-style feats wav _ from _ f bank.py : generatewavfromfbankusingwavenetvocoder文本处理filt.py3360

数据格式转换从data 2js on.sh download _ from _ Google _ drive.sh特性转换convert _ f bank.sh dump _ PCM.shwavscp到PCM波, feat _ to _ shape.sh generate _ wav.sh是来自fbank的与wav文件make_fbank.shmake_stft.sh模型相关的pack _ model.sh recog _ sh

net35离线安装,安装下载 pytest测试框架从入门到精通,pytest接口自动化测试框架