In the era of big data, more and more audio data is stored in the network, and audio-oriented automation and intelligent classification systems are in demand. Audio signal classification research is crucial to promote the development of such systems. In audio signal classification research, feature extraction, feature set optimization and classifier design are the three most important aspects. Through the research and analysis of existing audio related literatures, it is found that the focus of audio research mainly includes: research on different feature extraction methods in feature extraction; in the feature set optimization, research on different feature selection algorithms; In the design of the classifier, the classification performance of different classification algorithms is studied. This paper focuses on the above three aspects, and the completed work mainly includes the following parts: 1. Using different audio feature extraction methods, including time domain features, frequency domain features, cepstrum domain features and other features A total of 89 audio features of the category constitute the original audio feature set. For the feature set optimization problem, the feature selection algorithm based on Pearson correlation coefficient, the feature selection algorithm based on entropy weight method and the feature selection algorithm based on Relief algorithm are studied. An improved correlation coefficient feature selection algorithm is proposed. The validity and feasibility of the four feature selection algorithms are verified by experiments, and the advantages and disadvantages of the algorithms are compared. A decision tree based classifier, a K nearest neighbor classifier and a BP neural network based classifier are designed. The problem that BP neural network is easy to fall into local optimum is studied. A BP neural network classifier based on simulated annealing algorithm is designed. The original audio feature set is input into four kinds of audio devices respectively, and the speech and music classification is completed. Experiments, music genre classification experiments and music instrument classification experiments. The experimental results show that the average classification accuracy rates obtained by using improved BP classifier, traditional BP classifier, ID-3 decision tree classifier and K nearest neighbor classifier are 96.15%, 92.86%, 93.60% and 85.98%, respectively. The improved BP neural network classifier is used to classify the four optimized feature sets. The experimental results show that the classification results obtained by using the improved feature coefficient feature selection algorithm are the best in the speech and music. The average classification accuracy rates of classification experiments, music genre classification experiments, and musical instrument classification experiments were 97.78%, 92.69%, and 98.50%, respectively.
The audio expert is a very simple audio tool. Audio provides users with audio format conversion, audio merging and other functions. It can extract the user's need to retain part from an audio text file on the computer, and then make a new audio file. It is a good helper for users.
Audio Expert Features:
1. Music Format Conversion: You can convert between any MP3, WAV, WMA, AAC, AU, AIF, APE, VOC, FLAC, M4A, OGG and other popular audio formats.
2. Music segmentation: split a music file into several small music files, and support segmentation according to time length, size, average distribution manual and automatic.
3. Music interception: extract the part that needs to be collected from a piece of audio, and make it into an audio file.
4. Music Merging: Combine multiple different or identical music format files into one music file.
5. MP3 volume adjustment: adjust the volume of MP3 songs without any loss of sound quality. It is possible to analyze and adjust the volume of all MP3 songs to the same size before burning the music disc or before copying the song to the mobile phone. You don't have to adjust the volume of the player anymore when you listen to the song.
