Research On Copyright Protection Technology Based On MIDI Music Structure Characteristics

Information hiding science as a subject was formally born in 1996,the research of image,audio and video watermarking technology has gradually become the focus of the society.With the development of multimedia technology,the appearance of wav,mp3 and other audio formats has diversified the research of audio watermarking technology.MIDI audio first appeared as a digital instrument interface protocol,which greatly improved the speed of music transmission.Due to the advantages of small capacity and easy production,MIDI audio has been widely used in music production,game background and other aspects.MIDI audio records the playback speed and loudness of sound in the form of binary instructions,which is different from waveform audio.At present,compared with traditional audio watermarking researches,there are relatively few researches on MIDI audio watermarking at home and abroad,so the watermarking algorithm based on MIDI audio direction still needs to be further studied.This dissertation is based on MIDI audio structure characteristics,and puts forward two kinds of improved MIDI audio watermarking algorithm,further optimize the previous MIDI audio watermark algorithm,improve the MIDI watermark embedding capacity and security,and two kinds of suitable for MIDI audio quality evaluation objective method and put forward the objective audio quality assessment model based on absolute information entropy.At first,a double threshold audio watermarking algorithm based on MIDI time difference and intensity value is proposed.This algorithm improved the MIDI audio LSB algorithm of single parameter,the H-K melody extraction algorithm and time double threshold algorithm embedded the fluctuation,according to the appropriate threshold value time LSB embedding strength parameters,and achieve the goal of ascension embedding capacity,respectively the Logistic of watermark information and carrier and Arnold scrambling processing,achieve the goal of strengthening safety.Secondly,a MIDI audio watermarking algorithm based on DCT domain is proposed.In this algorithm,the pitch is considered as one of the attributes of sound,and the pitch is.determined by the fundamental frequency of the sound object,so the MIDI audio pitch is converted to the fundamental frequency.DCT transform can produce DC component and AC components,due to the absolute value of the DC component is greater than AC component value,therefore,this dissertation combines the watermark information with the DC component to obtain the pitch of the density,the difference between the densified pitch and the original pitch is embedded into the MIDI audio carrier as a new watermark.The algorithm combines the spatial domain with the frequency domain,which greatly improves the security of watermarking.Finally,in order to evaluate the transparency of the algorithm,the subjective quality evaluation method and the objective quality evaluation method are used respectively to evaluate the quality of MIDI audio containing watermark,and the absolute information entropy model is proposed to evaluate the audio quality objectively.In order to verify the anti-aggression of the algorithm,four conventional attack methods were selected for the attack experiment.Experimental results show that the watermark channel capacity of the MIDI audio dual threshold watermarking algorithm proposed in this paper can reach 6%-13%.Compared with the traditional LSB watermarking algorithm,the watermark embedding capacity is improved by about 3%,and the two algorithms proposed in this dissertation can resist external attacks to a certain extent.The two objective quality evaluation results are consistent with the subjective evaluation results,and the results of the absolute information entropy evaluation model are also consistent with the subjective quality evaluation results,indicating that the algorithm in this thesis has good transparency.