ganabathi post is totally unrelated.

Sound energy haversting is possible, but is also inpractical. In cellphone, most of the energy is used to transmit the vocal information through radiowave, while only small amount is used to in microphone, speaker and LCD. This is why talktime is waaay shorter (due to high transmission energy), than standby and music playback time.

Assuming ideal condition with no energy loss, meaning the sound energy from speaker and human voice is 100% transfer into electric energy, it would probably prolong our talktime by a tiny bit, which is why I say inpractical.

Wave energy is measured in decibel (db). Human voice probably generate some miliwatt or microwatt of energy, hardly usable.

Piezoelectric is used as sensor, not energy harvesting. They are used in various application in favor of last generation sensor technology because of the size, and superior electric characteristic. Their output signal is normally very small, and often couple with an amplifier to magnify the value so that it is readable by other devices.

P.S.: Some technical information sharing: