From 1622655542d0481c4fddb0ca2049c6c2c83f8e3a Mon Sep 17 00:00:00 2001
From: yemaozi88 <428968@gmail.com>
Date: Thu, 10 Jan 2019 23:39:02 +0100
Subject: [PATCH] functions are added to perform forced_alignment using
 novoapi. results can be written in novo70 or IPA.

---
 .gitignore                                    |   1 +
 .vs/acoustic_model/v15/.suo                   | Bin 83968 -> 84480 bytes
 .../__pycache__/defaultfiles.cpython-36.pyc   | Bin 997 -> 1099 bytes
 acoustic_model/defaultfiles.py                |  10 ++-
 acoustic_model/htk_vs_kaldi.py                |   2 +-
 acoustic_model/novoapi_forced_alignment.py    |  71 +++---------------
 acoustic_model/novoapi_functions.py           |  39 +++++++++-
 7 files changed, 56 insertions(+), 67 deletions(-)

diff --git a/.gitignore b/.gitignore
index b1d3894..520e3c4 100644
--- a/.gitignore
+++ b/.gitignore
@@ -3,6 +3,7 @@
 
 ## important ##
 .acoustic_model/forced_alignment_novo.py
+.acoustic_model/novoapi_functions.py
 
 # User-specific files
 *.suo
diff --git a/.vs/acoustic_model/v15/.suo b/.vs/acoustic_model/v15/.suo
index 7f32711d9e4ad6d5867d970c56cca4af30ba1635..9fe1f2136ee18d47f8e6c82b824e73835098274b 100644
GIT binary patch
delta 4103
zcmc&%3s98T75>l9?y@TjF3L+lSRNu$9=ou-4B#Tj!zvG77)u~L6-7XK2&k~uYLaSH
z_BgI_T3c7!Xk#4NzsjUDt*cNc@r9YOnKYu+#z_?)i76Uj61L~=E{4)N8plpA^Ks5S
z_nv#tf6lq*-qr0A>voHGC4qluaun~P<c@Y~j$DB>j#@Uh@~serkNC~}9l%zHR!K97
zGn6>&;N2cXti|3m#K(>4N7#8!qB1)lK4#o|41d7RyAqAD^CF^Q1UbQ%@Ez^rM>K#C
zN|?cM?OX^k6A6b1QwdWDQG{uPXu^*O%)S(&Mms-?s0(3)<OxKR>~c3dW%F6n+WUb-
z(+N_0pN(Y@ui`vxi)T_Ni!jeV0TnVOX34_EMI>m%7FlKh(_yWiBy3;S*4d}W5^d|q
z_Y`d+2ET<m7qh$L<L8NhW1uwxqn>AukPw*X*~gE1o^9w%eda;dp5=KqXVmjfq{sT>
zJkL$A$=fXoV#a%(Dg19eKk9gj<4f>74$8{aUO?$X@Fp;i`4XK<P!jy@GA0ik<vUC2
zK`(EBR&12FLkm8l|FziY+%7LBpENS>Q`1i<e2Gg-Y)cEnbFu-uc{RVD!=}vnGFV7k
zEg|5J_d~^QWa96Ip~;TCJ||1W@|}sXA!m_@9RzD<ASahCfDO5M(qs~{47Vx|5VKJj
za<_nFGOfdbFyeX^r2fRR(`B6_p#dolmAtFCmHSPfcjk{3TXx<(^46Bhg?PL)4^KDE
z!rrT5+_El4wwZLMEv~Ovi@GwMVUir84RKN6B53_?fS~lg0a}i`O3OW@%)L$YF&(p8
zM8Ba7J43qc{GTGia*DMT`D+t3xw#eP2??<=QL8lZt%i;$nC^P~mUzCP@vJPkzR@z$
zbD)gt$e)bc9>glHe-JwRSjLqzaBIrBGwM6@gv$?9U<wxT(x1|@TL~x+S<z9hkOdQe
zkTAV7zaHewaNCv^3(AXDIJ4njv0=hUXGuGl!UB0bG|3lJBL4tI!ce3m=o1>CS!hhy
z21XUHpP42DR{B_3vqfNLo|Kg*R(_ZOUc*N-+o2E}U)g5Vh)hXoAOp6k>WN13c9X~8
zXee;R=g%x^!ho+FOv-L+4xow`34;L*AZO{x!0HY*2dP3?r<yve2{i=PuOqsKu$C}d
zj_`7juQGtV&L*_cKs(`Q1nOGGxykAWtD9(iMCd45&X4=}OY{-X{n_|Ib|$uGOP~Q&
zP0^^!GGkEL9PC`4jIUo);D)A1Q@D~BlsPxRmOV;$kUExdbzBu!jH&kmJR^y3;a08j
zU%b9>*O!kvYCFAy=XO}*zy0qT*o&#t))<(U*s6tZ7x;wi+*tj|uEDs%gOft$@!x&%
z27SttV9zJMM#OR7ict5t?!U)GpJx_T*H=`K%@H_|7mdYbGB>u0HP!Rh)>c+kS4=Cc
zuSIRXP98}EtT?kGih13B4q=>BxSv0nH!kIkrSSCgqEAV~?pKmgmy#}Jp<qRo<&7|~
z))xfDg0H}s&NH`Y4012<bKTukwa4qVoo8EM%lRqt*9GPTp=sXF;pq#dP$E@o0aIH%
zu=a91wk@CTq)rG|ZOAC04_r?ln1po~#irSpL2+^WyT8n@JO0m=ruiRza`$g-cq2`z
zj|xPid!V#{OncRCp2U+<=>mttk)deJ2@$c&$67ahNd5pjVx#cavA&q!97I%U$5eZN
zifM@pphrO%6@m?<c-0BD1!n!+LEcG>!*fR*nK!8kmNei^<vdKvbl|v~bWt59=?OYo
z3pfm}n2SXf#+Ps0_jd4+oG`pl!W$Ry`eofRZYk}aPp}?zU7H98mt7FCn-O)f;bL~M
zI2?NQ+%qOx&oC)j8@9H@IH}`CgR#jP3{{T|YdYn)b=`DlHAPCTwuR~qHVQTRbvjh{
z_@E_k8Z=|7|15p=haxpCXSHjfx>w9CW<6#UPvPjpp;f(4nScYUJw=I{=mcS6%|n=2
zUQt=nSlv+Eu(~cfUXyMW8csVyqLA3^2#y;!sF|(!-txux=DJ|P+@j@;dWo@(H|`N(
z^yVI+XK#{|aR+ZYbRL{PA|q&fEB0aP5_GDgl2psBWD<1wjUS#nT!1$ZdjN<0krY99
z#LOFi!K3!54$bdt51Nls&T72S?_MYdzEvmG?f;463oNR27{3}cA{_q|51i_Ov*>ze
z8`>Q1hVNYUN5z@#Slk^iRG4dc+|#|1s@8K)zNU8OW=AYsG*d-?>thn)%SzP=<q>h}
z*q8{7IyNpMF*-IrBC(__rZO>F6<=Cfu@M6<{f_PH;Z{@IR9jM4Rb1IvQ`S&bTT>rd
z*D@bJ{wl$4o)aH7?@4z@Z;tyr<Xc+{Ftb}LoRE&Gr?-<rnlR_Q6kOR$+t$?PRK{S@
zg}>sai`j%tsP0p-sV{Cc_FwwincX_0EgWxtw)^|-XO7B=&LK=6v;9`||A(V<<C7Ty
z=Jj}?Z#VNjrhW0bD|_E_g%90}AD%eHp1OX*{lCnRSf5;89PqYxuw6S@@l4e)k>Jr+
z<shfG21muj+OYA?Hhk~PKMGeqO9Ry2Hsk)_5De+}LAYZ%=-J296+tfRx$}yYTiVOs
z!yIPzC!)DO6c6`%>A%(*?*haNaL)pmdrc|CU(bSn-(i1Wz<+)T-tS#De!J(}>I8tr
z0VmySK;3P(t~YsTbucKRAfzi)3ZKGB2^<q)pvBj4BNCQaFT}CesEh3y<v8|pM9<mJ
zR%_R6S5SzsB+bp>HUaj%kp7>fCzY@Ox)%7rG0C1_NP>L^aD%azvf*tvm?~yJANLwQ
zc7Rwz-UQGZ76d{6^UY1iVU=Ikl@PcNU8WGom|zHW25&=F7-YaM+lsmw@S))`2@Iv-
z;L&9Yg$${|5Cs80@acFm3CtjO65FnH8T}H%E!morv%$<eZP3Jw7Hh-#ba({6rN_rC
M?f>J2`@_xr7y8VN*#H0l

delta 3648
zcmcJR3s98T6@d3#c7cUOmM$!4c`S&!5k!^+St8<(P*lJatYD^tK|o#>Sqh6mMERpp
z+c;6!V_Y2*tINcuHpwVIEvcQ6RUCDqjWYE?d_<?sbSBo!Kt@NxDBE+FU5bfG?d#5b
zd(J)g_1|;O-h0J83b8}ckq+wP3+A#Z1?scC*N$9)#WI=9k&msc6*TyeZDmmmTfJJn
z+sQr>iKAZAijqlnPAUuCbrVUYTMwPoC(*6TNlqhXy6e9pnMni@y3RsPyFV#Y2+!QJ
z-1;GsDq^a;&f5*7hq&t+l26DMMKXy9CSD}w5~+k|zA)0`h-_j$F@xZ3W8)plor~Ot
zG?G7Z>qe3b-1-udONm?}kH{yM5ib!xCJKl`;wJ>3a|OvFqRm||?<|aquAs^qqQ-3;
zU)<@erTSXJOs-J=oRsy%2CC>^C8dSn-`PrXqg&rZayG$d5J_5D-IQbI_tsvgIL@Q`
zv*OrHb{<EcnD%f|BZ*kTo8Z&%_~{8AKVQO+;PK<`<;4@fK(f6<aGNWBq-%)?B7op-
z=50Kj@dW>L9?lntg9L@M-4)Mhs-zH}c>a)dKG0Jp$J1krp+4!~=;w)NhTAT<W!9vA
zIg|9|vd-lRxI_`JXd-JVZcr?a<__m;@Q<~S&V9R?nDBpRStyPu*L1E4S407As<SLh
zX&@zuSB{sZf>mXKjo7SehgN(@zYW+N(5_lZbL4Q3QnC7#{F}r5UZG**!=<sQXTg`d
zipL3e9oH+oX=z{SYmE=<4jTbE|8(i`)bmgG@^|&3Ayg!16fkRgKlbYk(t~sr*p?@N
zAPJFoLDFS(0k>l!lS?BDEubo)5qwS#wqz#i&19$}))DoS8hOGt65N|SSMhRYxW+S*
zJCPqCuao_7Rhndqy$G<KQX9i&mV`{F7CS*pOq7P>g|7nNkq@6k2Ci`^#Cb~@vp)wZ
zCVvc27e5H~(%11mU{8E%NA_g7HNgTKq~@GYK<Ua~o}=7fXh|yY<>v@bOm2Wz@L^Ir
ztiopFP83Z#NoSf)J}@poF-^3VWGR53A}FTBqot(4m*+P>X82I)L5T{wGx{f7=EMI{
zRBVL_V6>gp!4yfT=>YX<KVzR++tNAuo-;Y^#9KS?ZmrI~nsxsm3zWdm3jUAbGqOa$
z^_pea&kRw)6xbd<8<pCM(skxFm31W*=K8AI>V(?*S{%sJV_Tlmc5wkD`x;W_l^Tu4
zR1|X^U_fPVw##k{5g;7HbK|V_BkYkJN9)IY_;#p_f8$45&;_kc4c|SGxL^L*r|PP4
z?A<#u3XHVA{0*?b#H=eEK5aQ=bxoC3*jGIVhgT#`OO7|p8@KU4$&8|=%Ep35b5(6!
zB??8NiZ2jIxZJdQAd0ofq%Vs!%-+Jf>!yGrg2`lq=)Z0b3hNy9mzlM#UkSxDppbCI
zNbHc~NZUm@KSa3am98A1iVO{cMp6qU|7Bw8o{^ZcEi^yRzpZ>7mA<C4!B@#s^Dsf_
zxDl+`Bc6QwDZmUobvYh$^X9`g@=Pd-5n^{E1MfEKe$(IDr1cia3p;!;xG5Lwo2=bc
z0$|}^GVsmnS>8#h@diT*w&!Kx`-5t>G4VgjSL>GZuEcz><&qNkJ;M(gmj@qg3zetR
z3R2+qVdyqxJ%2`rgQQ+P1lTFgU`7*e%G-*4qJSM|wAR-|To_?e^4o_8y#+I)vL(eE
z^8p%BsMg}8%X3jB&eE9D;*-Yn!(=j5B+oI<tCAc$h40VzfXVp|_}ZQjN!xJ`zMZ%T
zZ!bo%Q!N!+kAl6Jrxorw(Bm?l*_RIgI-OJ#=|nsEP9OR+`UBpg9#kaf0bqe^Dyzzx
z>nx>~bqy1~#Z_-ACFQP13j020f7`G7qYDt=q!BdMMLur+KyN?D@W=<TUS!0D!(F)N
z#7aDMR7dwKzgo$-^tY*!@JGR`%W(`)JSJew{syVxPzOK`f@2WQV%fEwn0Qf%Fa9YQ
zPdGMr_gqkXYn}Y(U*J3D{i~$!uV+|!x*zvmxD6^C_-vQ;suLGoPsb;h_0MSSzM2O?
z+g!WgKL*Ec`enH4@~5yAhfkT%JfLQ?mv~e>KVU-Pnm<lw!7Tfm+lqo<G|&bw{dpPD
z5yaDZycit(#RSszPc;mkmzD+b>u0<e9Q?`&`#&!Ef;V<J6nOAV0G_<FXhM(pBuo0u
z-yDo(pQ7-zcip?maLJ9uQpC+P3Syzyb?-Kl4h_x(9X|<WtA|5Mhkw_{AHTfN14!oO
z(kTABV+^Rh{Y%VD4#UW=YEZZ@+P?&J+|Q)55)a2t$Eu-eJ}+P5uQ3_!7)nENNFQdI
zxE2zsYb~|a_2$M(-oZW$kRriXPLPUk)vzGGl&A3LRTKH@+}=#+QH@_322;Yk)CzPA
z3AQ^4Fe~U62j&M}JvV=4h${N(jdNF!-v`*YKSi@iT~Sk6@w3u;^LlexLv2FChVP59
zw%oHQN&aL|5H4RW!ot7npXo~}YVg<3;jb@lo%rAR%Kn&fca1u|dVsNBCWrUs%qnv)
z(Aqnsgp<tH7z_Gdy*F%UPqhWYNneb45^hubL2<8619!dm>fl4$@=z$~9gBqf0Q+pw
zaj*eL6C-<LVj&)I%e!IxaS;1d%S?#%vy}s6*j`V7k8P)8p{loM7Fd1%X~f3I+Fors
z9EE^K^6@)XNe^@jT|dU(X8^EXNQb^>H#(UAXMfNy`Y`9pY507q?ScRg)sqY#pEi7v
Pk4J7K?iIlCd_DgG#QyPd

diff --git a/acoustic_model/__pycache__/defaultfiles.cpython-36.pyc b/acoustic_model/__pycache__/defaultfiles.cpython-36.pyc
index b61bd5463db8263f528a3d6a2ae13e365b1c713e..635aaaed776cc3945d38c00f15e5d070810f80f6 100644
GIT binary patch
delta 363
zcmaFLewstsn3tEUT-r2-gNuRTF#{4{1F{`}xcJUQ<@%&%rYPQ2hAgHm=2YGk0U(br
zMKFpVD#Di{1QZc~@_15&Q$$jjQ$(9tqXeO194TTc;?2xaLQuX?ibRTJid2d;P>pa3
zV=#lJ%)}psj8T)b7)^lWd`5S3vt+9n|GZR}qRg_?nDAhqnC!%yluX@{qSCzVnAG&*
zn6$*))R<z^nB@GTg3{uc;*!kV+|<0uPZ=xJZZRho7v170$jHx2Elw@D#gvy*#h;g7
zmTzvL2NTp3ncU3e2ej-GQ@l1eRCjzzW>J&?oRe6R2o~XoiIgXnP1a#noE*=r%6MyX
nE3=)N5-^B3SXfvDL|7R482Ol37=@rTkjKKp#K$PYB)|j!^~_>-

delta 226
zcmX@j@swTJn3tF9%umgjcuod}#|%h-4ajx?;^GSvmFvS&1X2W>S)zDT8M2tNm{WOD
z_)`R$nWOmNe4!NK6y_9>X4WWvs2E3zXo?t6jR2G{kRqNUks_HQ1ym!L!Whh;DLt8m
zv5@iB<T6GRAi0{+eX<r)Ia?HSVsTN_<mF8Aj6#!-Gx-6f#F*nJFJ+dQ+{7$2`5?0@
lW7Op5%yvq0K-Y1wu&@Y-urTs5@-eY63c+asMiC|fCIGy1FuMQ%

diff --git a/acoustic_model/defaultfiles.py b/acoustic_model/defaultfiles.py
index f464b9f..e9c4c96 100644
--- a/acoustic_model/defaultfiles.py
+++ b/acoustic_model/defaultfiles.py
@@ -36,8 +36,14 @@ fame_s5_dir     = os.path.join(fame_dir, 's5')
 fame_corpus_dir = os.path.join(fame_dir, 'corpus')
 
 experiments_dir = r'c:\OneDrive\Research\rug\experiments'
-stimmen_transcription_xlsx = os.path.join(experiments_dir, 'stimmen', 'data', 'Frisian Variants Picture Task Stimmen.xlsx')
-stimmen_data_dir           = os.path.join(experiments_dir, 'stimmen', 'data')
+stimmen_dir = os.path.join(experiments_dir, 'stimmen')
+stimmen_data_dir = os.path.join(stimmen_dir, 'data')
+# 44.1 kHz
+#stimmen_wav_dir  = os.path.join(stimmen_dir, 'wav')
+# 16 kHz
+stimmen_wav_dir  = r'c:\OneDrive\WSL\kaldi-trunk\egs\fame\s5\corpus\stimmen'
+
+stimmen_transcription_xlsx = os.path.join(stimmen_data_dir, 'Frisian Variants Picture Task Stimmen.xlsx')
 phonelist_friesian_txt     = os.path.join(experiments_dir, 'friesian', 'acoustic_model', 'config', 'phonelist_friesian.txt')
 
 novo_api_dir = os.path.join(WSL_dir, 'python-novo-api', 'novoapi')
diff --git a/acoustic_model/htk_vs_kaldi.py b/acoustic_model/htk_vs_kaldi.py
index f4dd82a..ca7f6af 100644
--- a/acoustic_model/htk_vs_kaldi.py
+++ b/acoustic_model/htk_vs_kaldi.py
@@ -256,7 +256,7 @@ if make_kaldi_lexicon_txt:
         #        f.write("{0},{1}\n".format(key,c[key]))
 
         for key, value in c.most_common(option_num):
-            # make possible pronounciation variant list.
+            # make possible pronunciation variant list.
             pronvar_list = am_func.fame_pronunciation_variant(key)
 
             for pronvar_ in pronvar_list:
diff --git a/acoustic_model/novoapi_forced_alignment.py b/acoustic_model/novoapi_forced_alignment.py
index 93b6a73..932d7c1 100644
--- a/acoustic_model/novoapi_forced_alignment.py
+++ b/acoustic_model/novoapi_forced_alignment.py
@@ -37,11 +37,15 @@
 # Aki Kunikoshi
 # 428968@gmail.com
 #
+import os
+os.chdir(r'C:\Users\Aki\source\repos\acoustic_model\acoustic_model')
+
 import argparse
 import json
 
 from novoapi.backend import session
 import novoapi_functions
+import defaultfiles as default
 
 # username / password cannot be passed as artuments...
 p = argparse.ArgumentParser()
@@ -51,68 +55,11 @@ p.add_argument("--user", default='martijn.wieling')
 p.add_argument("--password", default='fa0Thaic')
 args = p.parse_args()
 
-wav_file = 'c:\\OneDrive\\WSL\\test\\onetwothree.wav'
-
-rec = session.Recognizer(grammar_version="1.0", lang="nl", snodeid=101, user=args.user, password=args.password, keepopen=True) # , modeldir=modeldir)
-grammar = {
-  "type": "confusion_network",
-  "version": "1.0",
-  "data": {
-	"kind": "sequence",
-	"elements": [{
-		"kind": "word",
-		"pronunciation": [{
-			"phones": ["wv",
-			  "a1",
-			  "n"],
-			"id": 0
-		  },
-		  {
-			"phones": ["wv",
-			  "uh1",
-			  "n"],
-			"id": 1
-		  }],
-		"label": "one"
-	  },
-	  {
-		"kind": "word",
-		"pronunciation": [{
-			"phones": ["t",
-			  "uw1"],
-			"id": 0
-		  }],
-		"label": "two"
-	  },
-	  {
-		"kind": "word",
-		"pronunciation": [{
-			"phones": ["t",
-			  "r",
-			  "iy1"],
-			"id": 0
-		  },
-		  {
-			"phones": ["s",
-			  "r",
-			  "iy1"],
-			"id": 1
-		  }],
-		"label": "three"
-	  }]
-  },
-  "return_objects": ["grammar"],
-  "phoneset": "novo70"
-}
-
-res = rec.setgrammar(grammar)
-#print "Set grammar result", res
-
-#res = rec.recognize_wav("test/onetwothree.wav")
-res = rec.recognize_wav(wav_file)
-#print "Recognition result:", json.dumps(res.export(), indent=4)
-
+#wav_file = 'c:\\OneDrive\\WSL\\test\\onetwothree.wav'
+wav_file = os.path.join(default.stimmen_wav_dir, 'pg_pauw_2206_0fjd8.wav')
 # list of the pronunciation for each words
 word = 'pauw'
 pronunciation_ipa = ['pau', 'pɑu']
-grammar = novoapi_functions.make_grammar(word, pronunciation_ipa)
\ No newline at end of file
+
+result = novoapi_functions.forced_alignment(wav_file, word, pronunciation_ipa)
+pronunciation_ipa, pronunciation_novo70, llh = novoapi_functions.result2pronunciation(result, word)
\ No newline at end of file
diff --git a/acoustic_model/novoapi_functions.py b/acoustic_model/novoapi_functions.py
index 0bdb324..0ab6aa8 100644
--- a/acoustic_model/novoapi_functions.py
+++ b/acoustic_model/novoapi_functions.py
@@ -1,7 +1,14 @@
+## this script should be used only by Aki Kunikoshi.
+
 import numpy as np
+import argparse
+import json
+
+from novoapi.backend import session
 
 import defaultfiles as default
 
+
 def load_phonset():
 	translation_key_ipa2novo70 = dict()
 	translation_key_novo702ipa = dict()
@@ -112,7 +119,7 @@ def make_grammar(word, pronunciation_ipa):
 
 	grammer_data_elements0_pronunciation = []
 	for id, ipa in enumerate(pronunciation_ipa):
-		novo70 = novoapi_functions.ipa2novo70(ipa)
+		novo70 = ipa2novo70(ipa)
 		grammer_data_elements0_pronunciation.append({
 			"phones": novo70.split(),
 			"id": id
@@ -135,4 +142,32 @@ def make_grammar(word, pronunciation_ipa):
 		"phoneset": "novo70"
 		}
 
-	return grammar
\ No newline at end of file
+	return grammar
+
+
+def forced_alignment(wav_file, word, pronunciation_ipa):
+	### IMPORTANT ###
+	# because of this function, this script should not be uploaded / shared.
+
+	# username / password cannot be passed as artuments...
+	p = argparse.ArgumentParser()
+	p.add_argument("--user", default='martijn.wieling')
+	p.add_argument("--password", default='fa0Thaic')
+	args = p.parse_args()
+	
+	rec = session.Recognizer(grammar_version="1.0", lang="nl", snodeid=101, user=args.user, password=args.password, keepopen=True) # , modeldir=modeldir)
+
+	grammar = make_grammar(word, pronunciation_ipa)
+	result = rec.setgrammar(grammar)
+	#print "Set grammar result", res
+	result = rec.recognize_wav(wav_file)
+	return result.export()
+
+
+def result2pronunciation(result, word):
+	result_ = [result[i] for i in range(len(result)) if result[i]['label'] == word] 
+	llh = result_[0]['llh']
+	phones = result_[0]['phones']
+	pronunciation_novo70 = [phone['label'] for phone in phones]
+	pronunciation_ipa = [novo702ipa(phone) for phone in pronunciation_novo70]
+	return pronunciation_ipa, pronunciation_novo70, llh
\ No newline at end of file