Simultaneous observations of Jupiter in the 3.5-4 μm range with a camera on the NASA IRTF and a Fourier transform spectrometer on the CFHT are analyzed to retrieve the spectral, spatial, and temporal characteristics of the polar emissions. At 3.50μm, a uniform planetary background is observed, with a spatial extent similar to the polar haze observed at 8900 Å and 2 μm. After subtraction of this continuum, pure H 3+ emission appears in the corrected images at 3.53 μm. Maximum emission is observed in a limited latitudinal range, between 60° and 70° latitude, both north and south. The longitudinal distribution of emission exhibits strong maxima, which are called "auroral spots", from which most of the emission originates. In the north, two spots are observed, separated by about 60°, with a narrow bridge between them, forming a weaker "arc" of emission, centered at about 180° of System III longitude. At longitudes outside this arc, the emission is weaker by at least a factor of 8. The spots are roughly fixed in System III longitude. In the south, the description is more complex. The emission is also mostly confined to auroral spots, but the longitude of the spots is found to vary temporally. Two spots are observed in a time series covering half a Jovian rotation. Compared to the north, the emission is mostly confined to a broader longitude range, -20° ≤ λIII ≤ 140°. Within this range, a narrow emission peak is observed to shift with time toward greater longitude, at a rate slightly smaller than the Jovian rotation rate. It is observed at almost constant local time, about 20° to 35° eastward of central meridian. The intensity of the spot varies continuously, with an absolute maximum for longitudes 60° ≤ λIII ≤ 90°. The spatial characteristics of the H 3+ emission are compared to the Voyager UVS Ly α observations, to thermal infrared auroral observations, and to recent models of auroral emissions. The pairs of northern hot spot peaks are accurately located on the polar edge of the "UVS auroral oval." The southern hot spot peaks are located, within the error bars, on the theoretical O 4L = 8 magnetic footprint. In System III longitude, the north peaks bracket the average locations of the UV peak and of the CH 4 IR hot spot. In the south, the apparent motion of the spot, as described above, can be interpreted as the convolution of an emission peak fixed in longitude at λIII ≈ 80°, close to the UV emission peak, with a modulation that is fixed in local time and maximum at about 30° from the central meridian. Comparison with the models suggests that within a pair of spots, each one could be related to different precipitating particles: electrons and ions.