Jump to Menu.

Jump to News.

Keio University - Science and Technology
Japanese
Sitemap
Outline
Faculty/Graduate School
Admissions
Study in Japan
Staff Search
Links
Access

HOME > Encouragement of Learning > What does your voice look like?

What does your voice look like?

SUGIYAMA Yukiko (Ph.D.)

My specialization is phonetics. Specifically, I’m interested in analyzing acoustic properties of speech sounds and investigating the mechanism by which we perceive speech. Suppose your friend says “ohayou (good morning),” to you. You will first hear “o,” and then you will hear “ha,” which is followed by “yo” and then by “u.” It sounds as if the four distinct sounds come into your ears one after another. Acoustically, however, speech is continuous with no clear boundaries. It is your brain that interprets that you hear those four separate sounds.

As an example, let’s take the acoustic properties of the speech “ame o nameru (eat candy).” As a first step, when you think about what speech segments it consists of, you realize that it consists of six vowels and four consonants:
/a/, /m/, /e/, /o/, /n/, /a/, /m/, /e/, /r/, /u/.
Now, let’s see what it looks like. Figure 1 shows waveforms of me saying “ame o nameru.” The waveforms represent pressure changes in the air caused by the words uttered from my mouth. While the amplitude is not uniform, it does not tell you where the consonants and vowels are. Figure 2 shows the spectrogram or the “voiceprint” of the same utterance. The darker an area is, the greater the energy around that frequency. The spectrogram does not show any breaks between the consonants and vowels. These two figures indicate that, at a physical level, one speech segment gradually changes into another; it does not abruptly change into another like we hear speech sounds. In other words, without even being aware of it, our brain achieves such a sophisticated feat to divide continuous, ever-changing speech into chunks that are linguistically meaningful as we perceive it. This is called “segmentation” in technical terms. Why is segmentation a feat? Without segmentation, we would have to think like, “this sound is something between /a/ and /m/,” or “this sounds like /m/, but it has a hint of /a/, so it’s not quite /m/.” If our perception worked this way, speech communication would be impossible.

Speech analysis used to require big, expensive equipment. Nowadays, however, you can download free software called Praat (it means to speak in Dutch) on your own computer and analyze speech sounds. What does your voice look like?

The URL to Praat’s website: http://www.fon.hum.uva.nl/praat/

Japanese version

↑PAGE TOP