As an indispensable woodwind instrument in the symphony orchestra, the clarinet, with its wide range and rich timbral variations, is regarded as “the instrument closest to the human voice”. In symphonic concerto, the clarinet plays an important role, not only in creating dramatic conflicts and interpreting the characteristics of woodwind instruments, but also in adding special colors to the music. This paper takes the relationship between clarinet technology and symphonic concerto in music education environment as the research object, and constructs a clarinet music classification model based on machine learning. The study adopts the feature extraction method combining Mel Frequency Cepstrum Coefficient (MFCC) and Perceptual Linear Prediction (PLP) to construct a selfconstructed symphonic concerto dataset, which contains audio of six instruments, including violin, viola, cello, oboe, clarinet, flute, etc., with a total of 1.25 GB of data, and a playback time of about 7.46 hours. Subsequently, the Improved Particle Swarm Optimization algorithm (IPSO) is proposed to optimize the classification model of Support Vector Machine (SVM) to achieve accurate recognition of clarinet music in symphony concertos. The experimental results show that the MFCC-PLP algorithm is significantly better than other feature extraction methods with the number of times the feature items are selected close to 51 times in both the training set and the test set.The IPSOSVM model achieves a classification accuracy of 99.09% for the clarinet music, and the mean of the overall music classification correctness is 97.43%, with a classification time of only 1.67 s. This method can be used to recognize the clarinet music in the symphony concerto, which is the most important music in the symphony concerto. The method provides an effective technical support for the intelligent recognition and performance optimization of clarinet music in symphonic concertos.
