Forensics/Analysis/First steps before analize the human voice
The goal of the recognition automatic speech (RAH) is to imitate the process of recognition which takes place the receiver in the comunication. There are several levels of recognition in a human process, and different automatic recognition systems implement all-some or only the most basic depending on what is it application and complexity.
Secure data removal
[edit | edit source]We must proceed to capture secure data for further treatment and analysis. The devices extract all information from a mobile device (address book, record calls, voice, pictures, text messages, etc.).
LPC (Linear Predictive Coding) analysis
[edit | edit source]The LPC analysis focuses as a process of separation between the fine structure of the spectrum (harmonic frequency distribution) and the spectral envelope (formed by the peaks corresponding to each element of the signal).
Histogram
[edit | edit source]Graphical representation of the fundamental frequency (F0) of the audio signal, its average value and the corresponding deviation. Statistical values also provide us with the percentage of time with voice and non-voice, and the intensity of the signal.
Double FFT analysis and transfer function
[edit | edit source]Through a mathematical calculation we can relate and compare 2 audio signals to establish similarities or differences in the domain of the frequency, amplitude and time. Very reliable procedure for locating small fragments, phrases or words that have been cut or replaced by others.
