hello everybody ,i am student in electronics for my final year project "restoration of clipped speech signal" i have to restore clipped sample speech that will be used for speech recognition.
1_ i want to know why the recorded speech signals recorded by a microphone are peak clipped ?
2_ are there any comercial software that perform the restoration

Clipping occurs when the output voltage of one device is too high for the input of the next device in the signal chain. The signal in the receiving device rises, along with the external applied signal, until it reaches its maximum, generally somewhere around the power supply voltage to its input stage. The applied signal may continue to rise after that but inside the receiving device there is no where else to go, it is just flat at maximum.

For many analogue devices (e.g. a power amplifier), distortion may increase considerably as the signal approaches maximum, i.e. the relationship between applied voltage and response to that applied voltage may become non-linear. Often that means the internal signal starts to rise much more slowly than the applied signal. Regardless, at some point it gets as high as it possibly can and results are as described above.

With a microphone, that overload might possibly occur in the microphone itself, I suppose. I'm not very familiar with microphone internals but there is generally some electronics in there to amplify the signal the diaphragm generates. If too much pressure is applied to the microphone diaphragm, the signal crated by the diaphragm movement may be too much for those electronics. However, my one decent microphone has a maximum input of 148 dB SPL for 1% T.H.D. (or 158 dB SPL with its 10dB pad engaged). Speech is very unlikely to be loud enough to overload that microphone, but some microphones may be less able.

When audio is digitizes by an ADC, there is always some maximum value that the ADC can handle, a value at which all the bits are used. Ideally, the analogue front-end to the ADC is linear well beyond that point, so no distortion occurs prior to clipping. At one level the recording is perfect, at the next tiny rise in voltage it is terrible -- clipped. The waveform goes completely flat at that maximum value and stays there until the input signal drops below the DAC's maximum.

Adobe Audition has a clipping restore function. So does the Sonic Foundry Noise Reduction plugin. Probably many other audio editors do also.

Your restoration project may be primarily aimed at demonstrating to you that no clipping restoration can be very good. One should always strive to avoid clipping. Most things being digital recorded these days, clipping is much more likely to take place in the analogue to digital step than anywhere prior to that. When digital clipping occurs, there is simply zero information in the recording about what the input was like during that time. You may be able to make your recording sound better but it is just not possible to make it what it should have been.