The rise of new energy technologies is transforming the construction machinery and heavy vehicle industries, with high-speed, precision electric drive gearboxes becoming crucial. In environments requiring high speeds and heavy loads, the absence of engine noise exposes the whistling noise issues in these gearboxes. Traditional noise analysis of gearboxes primarily collects noise signals and uses Campbell diagrams to analyze noise orders. However, these analyses lack specific identification and detailed investigation of whistling sources. To address this gap, we propose a multi-channel fusion method of acoustic and vibration signals to trace gearbox whistling sources. This method converts multi-channel signals from the time domain to the frequency domain using the short-time Fourier transform. It applies a prominence ratio to filter the spectral results, retaining the relevant whistling frequency bands. The processed spectra from each channel are categorized by dimensions and synthesized, combining both acoustic and vibration data. Multiplying the spectra at each frequency point yields a fused multi-channel spectrum, which is then converted from the frequency domain to the order domain to trace whistling sources. Experimental results confirm that our method can effectively and accurately trace the sources of whistling in electric drive gearboxes and identify the contributions of each source.