Music instrument recognition through machine learning typically converts sounds into spectrograms as input for classifiers. Many different types of spectrograms exist. This study explores the performance of various spectrogram techniques for musical instrument classification using deep learning and convolutional neural networks (CNNs). We compared six spectrogram methods: Short-Time Fourier Transform (STFT), Log-Mel Spectrogram, Mel-Frequency Cepstral Coefficients (MFCC), Chroma, Spectral Contrast, and Tonnetz. Additionally, we combined these techniques to examine if a fused approach enhances classification accuracy. Our results indicate that the combined spectrogram slightly improves overall accuracy. This research contributes to the field of music information retrieval by providing insights into the effectiveness of different spectrogram techniques and their potential combined application in future machine learning studies.