Friday, May 15, 2009

Busy Week

Baru lps abis jumpa SV. Alhamdulillah, lega...

Patutnya jumpa SV smlm, tp aku yg kansel sbb pgnya ada session 'Effective Poster Presentation' sampai pkl 12.30. Pastu disambung ngan Postgrad Society Lunch Exchange (ps: mknan sgt sedap, n good way to socialize). Ptg nya plak aku volunteer jd bahan ujikaji utk ada psychology test ni, kena pki EEG cap gitu mcm ni. Gmbr aku pki cap tu ada, tp x leh share sbb nmpk rambut :) Before balik sempat main squash, so risau kalo2 experiment tu over-run kang, naya je SV aku tggu.

Jdnya terpaksa la aku gi school hari Jumaat. I am indebted kpd Hubs yg sanggup keja overtime hari ni jaga Aisyah, sbb nya Jumaat patut aku x gi school, leceh, Hubs nk gi masjid etc. So around 1 pm, I met them dkt simpang masjid n bwk Aisyah balik lab.



Gini la rupa Aisyah masa di lab aku. Main btg reket squash aku. Dan aku dlm background yg tgh prepare2 apa yg patut before SV masuk. X de org nk letak gmbr ni kat Kosmo! ke? Buat artikel tajuk 'Ibu, Isteri, Pelajar... Di Perantauan' ke plak? Kih kih... Just kidding... Saja je aku nak nyakat cik Yzma, yg kini glamer you, masuk Kosmo. Petikan akhbar tu di sini.

Anyway, not long after Hubs n Aisyah beredar, perjumpaan ku bermula. Perjumpaan ini in my standards, bagus and membuahkan hasil. Tolong la jgn label aku riak, bkn niatku. Just that before this, entry mcm byk geram, down, mencarut, etc, asyik negative ajo, sampai my parents kerisauan di Msia. Alah, Umi & Ayah ni, mcm x kenal anak perempuan korg yg sorg ni mmg sgt2 over dramatic. hehe... So skang letak benda yg positif plak...

Perbincangan td kejap je. Tp bermakna. Sblm ni duk buat manual segmentation, by ear je. Nak test market, kata org. So hasil perbincangan last perjumpaan, kena find a way to segment audio automatically, and preferably put some kind of fade-in, fade-out effect utk enveloping pada setiap segments supaya tidak ada clicking effect.

So antara masa yg lps dan masa yg skang ni adalah dihabiskan dgn mencari jalan la utk wat benda tu. Ada time rasa bosan and ragu2, betul ke aku buat ni, coz it felt like the days go by mcm sia2, coz searching endlessly over the Internet ni byk makan masa, and belom tentu membuahkan hasil. I tried to find tools aje, spt yg byk made available kalo berplatformkan Linux, sbb yours truly malas nak compile and debug and correct code org. Bkn contribution pon lg, just testing the water je dulu, so malas nak get all dirty. Teruk2. Of course, ada aje software yg x free, berlambak, tp sapa nak byr?

Anyway, dlm mencari2 tu ada la jumpa mcm2. Some downright useless, some appeared like there could be something, rupanya useless, some at one point useful, pastu sbb author dia dah tarik dari Internet so jd useless la utk aku. Tp byk jg yg useful.

So antara yg useful tu la present kat SV. And diskas the pros and cons of each. Kot2 la ada reader interested, bole search under Adintool, Audacity Beat Finder, AudioSeg, Multi-Class Audio Segmentation Tool (MATLAB), WavSplit, Slice Audio File Splitter (WINDOWS) and Aubio (yg x buh bracket tu by default Linux). Of course la masa wat pemilihan tu kena la bg summary of what parameteres they optimized, what algorithm, pitch detection mode, blah blah they used. Di sini la kena tunjuk mcm kita bkn search aimlessly, so kena la presentkan benda2 yg dah baca dari theses diorg tuh.

Actually, in my humble opinion, Aubio is quite good at doing the job that I need to do. Dia segment mengikut onset yg ada pd audio file yg dinyatakan. Aubio ni hasil PhD work Paul Brossier, dari Queen Mary beberapa tahun yg lps. In fact, SV aku ni yg jd external examiner dia, so dia pon teringat balik how relevant his work could be to what I am doing. Excited my SV. Dia pon dah x ingat la tu hehe... Plus, the thesis was quite easy to follow step-by-step, kira mmg bagus la penulisan mamat ni. My SV pon puji2 lg, 'he was a good student, and his thesis was very clear. Maybe this can inspire you to write like him'. Woh!

Aku ni mmg slalu insecure, so confessed la kat dia, 'rasa mcm blur2 la wat benda ni'. Dia jawab, 'Ah, yes, when you are doing a research you can sometimes feel at lost'. Aku jawab la balik, 'I just hope I am not drowning in my own research', to which he replied, 'No, no, no, you are not drowning. This is good.'

Oh! Music to my ears. Btw, insert gosip sket, did I tell you my SV is Brazilian, bkn British? Makanya, accent dia adalah sgt chumil. Spt, 'Nu nu nu, yu er not drew ning. Dis is guuud...'

Serius comel kan?

So gitu le lebih kurg nya meeting tersebut. Next goal, autosegment kan all the dataset (lagi bagus kalo leh dpt yg dari MIREX - pergh!) and re-start concatenation. Next meeting in two weeks. We'll see what the result of concatenation is....

Oops, dah dkt pkl 5 ptg. Seeing that hari ni patutnya hari cuti aku, nk balik ah. Have a nice weekend, y'all!

No comments: