Discrimination of Audio Signals Using Time-Frequency Distributions

Humans can discriminate speech from music easily in their mind without any influence of the mixed music. Due to the new techniques of analysis and synthesis of speech signals, the musical signal processing has gained particular weight, and therefore, the classical sound analysis techniques are used in processing music signals. Music art has a long and distinguished history. It goes back to the time of Greek and is developed through centuries in both the musical instruments and melodies. The problem of audio signal classification serves as the fundamental step towards the rapid growth in audio data volume [1], [2], and [3]. There are many kinds of music such as: Classical, Rock, Pop, Disco, Jazz, Country, Latin, Electronic, Arabic, etc. [4], [5], [6], [7], and [8]. Audio signals change continuously and non-deterministically with time [12]. Consequently they are usually characterized as time averages, and their relative amplitude and frequency contents can be easily specified. As an example, speech and music typically have strong low-frequency energy and progressively weaker high-frequency content [9], [10], [11], and [12]. The maximum frequency, fmax, of an audio signal varies according to audio signal kind; fmax equals 22 KHz in CD quality recording, 11 KHz in FM stereo broadcasting, 6 KHz in stereo or multi-loudspeaker recording, 5 KHz in mono-loudspeaker recording, and 4 kHz in the traditional telephone transmitting quality. A generalized frequency spectrum for audio signal is shown in Figure 1, and audio signals can be classified into the following classes: 1Speech signal compounded of single talker in specific time period. 2Completely music signal without any speech component. 3Mixture of single talker speech and background music. 4Songs; mixture of music with a singer voice. 5Singing without music. 6Abnormal music; uses acclaim cadence, single word cadence, human whistle sound, opposite reverberation or any non-music sound that been inserted as a basic tone of the music melody. These cadences cannot be generated by any of the ordinary musical instrument except modern Organ and mainly processed by a major help of computers. 7Speech signal compounded of two or more speakers talking simultaneously in a specific time period. A good algorithm for separating the sounds of two talkers taking simultaneously is using the cepstrum analysis. 8Non-speech and non-music signals: like car, motor, fan sounds, etc. 9Complex sound mixture like multi-speakers or multisingers with multi-music sources.