Joke Collection Website - Mood Talk - How to make the piano sound like human voice?

How to make the piano sound like human voice?

Pressing different keys can be regarded as a linear superposition of audio signals. The sound that the piano can pop up is a linear combination of 88 keys. Then from the frequency domain, the components of different frequencies are also linearly superimposed. Generally speaking, most people can distinguish the pitch difference of at least 25 notes, that is, 1/4 semitones. The frequency of human speech is about 300 Hz to 3400 Hz, which is equivalent to 42 semitones. For the sake of simplicity, when people listen to others, they are distinguishing the amplitudes of these 42*4 = 168 pitches with different frequencies. If you have a little knowledge of linear algebra, you will know that the linear combination of 88 keys cannot cover all possible combinations of 168 frequencies. It's like using only a linear combination of two three-dimensional vectors can't represent all three-dimensional vectors. Even if a certain combination of 88 key sounds can be very close to human voice in the range of 300 Hz to 3400 Hz, it is difficult to say what kind of performance is outside this frequency range, which is likely to change the auditory experience. In addition, this ingenious combination must be rare, otherwise it is easy to make voices when playing the piano. In view of this connection, someone abroad has made a robot, which can make the piano "speak" by controlling the keys through a computer. The tubing link is here (a ladder is needed): if you don't look at the subtitles, some simulated voices can be heard and some are difficult to distinguish. After all, if only 88 tones are used to cover all the combinations of 168 frequencies, then there must be a lot to approach. Nevertheless, it is generally feasible to simulate human voice with piano in video display. Initially, I took a course in signal processing. The content of an experiment is that a person records a passage first, and then uses MATLAB to carry out Fourier decomposition, leaving only a few frequency components with the largest amplitude.