2008年11月11日 星期二

要瘦?讓怎麼穿?


       前兩個月的Times上,有一篇很有趣的科學新聞,這篇所報導的內容,似乎顛覆了設計界長久以來的常識--穿橫條紋的衣服,會比穿直條紋的衣服,看起來要來得瘦。各位可以到這裡,看到原文。

        這篇文章所訪問到的專家Peter Thompson,是在視覺研究領域上非常有名的人物。每年在美國Florida所舉行的Vision sciences society年會,Peter的演講總會被放在最後壓軸,和他排在同一個時段的講者常常會自我調侃:「看來聽眾大多跑去Peter那裡了吧。」他的演講之所以受歡迎,除了Peter英式口音與特殊的幽默之外,主要的原因在於他會在投影片中,放上許多demo,而且絕對不是只有那種實驗室在用、枯燥無味的圖表,他會從生活中找到印證他科學發現的實例。

        Peter Thompson舉了一個古老的錯覺例子,這個錯覺是19世紀的科學家Hermann von Helmholtz發現的:在一個正方形中,放入等數目的線條,橫著放時,正方形會看起來比較窄,直得放,會看起來比較寬。如下圖所示:

       
(這個Blog圖片不加邊框,也許看不太出來,不過各位可以自己用小畫家試試看)

然而,這個錯覺並沒有受到設計界的重視,且反而發展出完全相反的常識--穿直條紋的衣服,看比來會比較瘦。似乎沒有人知道這樣的常識,是來自於什麼樣的理論基礎。而且Peter表示,時裝界的人,仍不願相信這個事實。

        設計界的朋友先不用著急,Peter表示也不是所有的衣著常識都是錯的。「黑色能讓人看起來比較瘦」這點是沒有錯的,如下圖:

       

黑色的圓形,會看起來比右邊白色的圓形要來得小。如果你覺得看不出來,或是效果很小,沒有關係,我們會在稍後談到這個問題,它與Peter最後想強調的重點很有關係。

        除了穿橫條紋會看起來比較瘦外,這篇報導少說了Peter所報告的一個結論--讓自己看起來更立體一點。這個建議便是他在會議演講中所提到的,當條紋出現在立體物體上時,上述的效果會更為顯著。筆者為了驗證這項結果,自己利用繪圖軟體製做了幾個衣服圖樣,並問問身邊的朋友覺得哪種圖樣讓筆者看起來最廋。以下是我所提供的圖樣:


(橫條紋服飾)


(直條服飾)


(要看起來更立體,細一點的條紋看起來距離比較遠)

        其實問過幾位朋友的感覺,條紋的增益效果並不明顯,而且有些人也誠實地表示,他們是根據「時尚常識」來回答的。然而,增加立體感的圖樣(第三張圖)的效果就較大了,有人表示:「好像胸肌很大。」或是:「肚子好像比較小。」由這個小小的實驗結果中我們可能看到,想要利用視覺錯覺來讓自己看起來較苗條,並不那麼務實,這也是這篇報導中的最後Peter要提醒大家的:「沒有條紋可以真正地消除你的肥胖。」各位朋友,還是花心思在運動、正常飲食上頭吧。

.
.

2008年3月28日 星期五

Device and Experiment



........In the last article, I just said that I got my first TvC function. But that was a pilot study, and I didn't control some environment factors like the environment luminance. I didn't calibrate the monitor before the pilot experiment. So after confirming the program worked, I began to do the calibration stuff. Unfortunately, it took me about a whole week. At the beginning, the measurement device, the lightmous, didn't work. After changing different monitor and PC, finally I found the offender, the USB keyboard. Because the driver of the lightmouse is the same with the USB mouse "and " USB keyboard, so if we plug the lightmouse and any one of the USB mouse or keyboard simultaneously, the Window system will get confused. Therefore, I replaced the new USB keyboard to an old PS2 keyboard.

........But the tragedy not end yet, the gamma table that I got was a little bit odd. The last 18 level of the gamma function was flat. So again, I change the monitors and computers. This time, I figured out the suspect was the video card then got a normal gamma function. And because I needed to use another video card, I had to use another computer. I installed the Psychophysics Toolbox of version 3.0 in the new one. Besides my boss, Dr. Chen, I am the only one voluntary pioneer to discover that. There are some routines changed in the version 3.0. The order of the arguments that should be passed into the Screen function, for example, is different:


Screen(win,'PutImage',...) become Screen('PutImage', win,...)


Therefore, all the old exp program are needed to be modified. And the biggest one is that, version 3.0 in Window system "CAN NOT" change the color lookup table!! Unless the lookup table we passed is a monotonically increasing lookup table. So if I want to present the colorful stimulus, the matrix of the image should be an n-by-m-by-3 one that each the third dimension indicates the intensity of different guns respectively. And also, we can't play Clut animation any longer...

........Anyway, yesterday I overcame all of the obstacles and everything finally went just right. Should I love this part in the research? It's quite hard to answer this question. Everyday I left the lab without figuring out the problem with frustrated mood, that is really bad. But yesterday I left the lab with hope and confidence, however, that is great. And a loss may turn out to be a gain.
.
.
.

2008年3月21日 星期五

My first TvC function



........Last week, I got my first TvC function by myself. On the original purpose, I was testing the experiment program and debugging. I used Quest procedure to find subject's threshold in that experiment and this was quite new for me. I am not familiar with the Quest procedure actually, but in general, the Quest is based on the Bayesain probability model. It will use the accuracy subject makes in certain difficulty level to predict the correct rate in more difficult question. So the difficulty of the next question is depended on your answer is correct or not in current question. In conventional, the threshold of a given pedestal stimulus is defined by the intensity which subject has 75% probability to see the test stimulus. According to the Quest procedure, it takes about 32~40 trials to determine that intensity, said just-noticeable-difference (jnd). As long as changing different pedestal stimulus, we can get different threshold for different pedestal stimulus, then we can draw a function that is changed along the intensity of the pedestal stimulus. This function we call the TvC function, means threshold versus pedestal contrast function. My TvC fuction looks like that:

we call the function with such kinds of shape "dipper function", just because its shape likes a dipper. If the TvC function has this shape, it implies that the our visual system responses to the stimulus in the way that just like the neuron system does. As the intensity increasing, the neuron activation will first be increased very fast, then the increment speed will slow down, and finally the activation will meet to the saturated level. Therefore, When the intensity is became stronger and stronger, the threshold will be decreased at the beginning, then gradually increased. This kinds of the response pattern results in the dipper-shaped function as the figure shows above.

........When this figure popped out, I was so exciting. After several hours programming and repeating hundreds of boring trials, finally I got the function with dipper shape. This experience was so impressive and awesome. The feeling quite likes that I got my first fMRI activation image, even stronger than that. Honestly, I really love this part!
. . .