This paper proposes multivariate multiscale Higuchi fractal dimension (MvmHFD) and its application to mechanical signals. Traditional fractal dimension (FD) can only analyze the complexity of a single time series at a particular scale. To characterize the complexity of multichannel time series, multivariate Higuchi fractal dimension (MvHFD) was introduced, and MvmHFD was proposed to analyze complexity at multiple scales. The effectiveness of MvHFD and MvmHFD was verified through simulated and real signal experiments. The simulation experiments tested the stability, computational efficiency, and signal separation performance of MvHFD and MvmHFD, while the real signal experiments tested the effect of MvmHFD on the recognition of multi-channel mechanical signals. The results showed that MvmHFD achieved a recognition rate of 100% for signals in three features, which was at least 17.2% higher than for other metrics. MvmHFD was found to have the best stability, computational efficiency, and signal discrimination capability. It was also shown that MvmHFD had the highest recognition rate for mechanical signals, especially when multiple features were used. The study concludes that MvmHFD is an effective method for analyzing multichannel mechanical signals.This paper proposes multivariate multiscale Higuchi fractal dimension (MvmHFD) and its application to mechanical signals. Traditional fractal dimension (FD) can only analyze the complexity of a single time series at a particular scale. To characterize the complexity of multichannel time series, multivariate Higuchi fractal dimension (MvHFD) was introduced, and MvmHFD was proposed to analyze complexity at multiple scales. The effectiveness of MvHFD and MvmHFD was verified through simulated and real signal experiments. The simulation experiments tested the stability, computational efficiency, and signal separation performance of MvHFD and MvmHFD, while the real signal experiments tested the effect of MvmHFD on the recognition of multi-channel mechanical signals. The results showed that MvmHFD achieved a recognition rate of 100% for signals in three features, which was at least 17.2% higher than for other metrics. MvmHFD was found to have the best stability, computational efficiency, and signal discrimination capability. It was also shown that MvmHFD had the highest recognition rate for mechanical signals, especially when multiple features were used. The study concludes that MvmHFD is an effective method for analyzing multichannel mechanical signals.